arXiv:astro-ph/0212531v1 23 Dec 2002

arX

iv:a

stro

-ph/

0212

531v

1 2

3 D

ec 2

002

Astronomy & Astrophysicsmanuscript no. new October 31, 2018(DOI: will be inserted by hand later)

The Andromeda Project. I. Deep HST-WFPC2 V,I photometry of 1 6fields toward the disk and the halo of the M31 galaxy.

Probing the stellar content and metallicity distribution. ⋆

M. Bellazzini1, C. Cacciari1, L. Federici1, F. Fusi Pecci1, and M. Rich2

1 Osservatorio Astronomico di Bologna, Via Ranzani 1, 40127,Bologna, ITALY2 Dept. of Physics and Astronomy, Division of Astronomy and Astrophysics, University of California, Los Angeles, CA 90095-

1562

Received ; accepted

Abstract.We have obtained HST-WFPC2 F555W and F814W photometry for 16fields in the vicinity of the luminous nearby spiralgalaxy M31, sampling the stellar content of the disk and the halo at different distances from the center, from∼ 20 to∼ 150arcmin (i.e.∼ 4.5 to 35 kpc), down to limiting V and I magnitudes of∼ 27.The Color-Magnitude diagrams (CMD) obtained for each field show the presence of complex stellar populations, includinganintermediate age/young population and older populations with a wide range of metallicity. Those fields superposed on the diskof M31 generally show a blue plume of stars which we identify with main sequence members. According to this interpretation,we find that the star formation rate over the last 0.5 Gyr has varied dramatically with location in the disk.The most evident feature of all the CMDs is a prominent Red Giant Branch (RGB) with a descending tip in the V band,characteristic of metallicity higher than 1/10 Solar. A red clump is clearly detected in all of the fields, and a weak blue horizontalbranch is frequently present.The metallicity distributions, obtained by comparison of the RGB stars with globular cluster templates, all show a long, albeitscantly populated, metal-poor tail and a main component peaking at [Fe/H] ∼ -0.6. The most noteworthy characteristic ofthe abundance distributions is their overall similarity inall the sampled fields, covering a wide range of environmentsandgalactocentric distances. Nevertheless, a few interesting differences and trends emerge from the general uniformity of themetallicity distributions. For example, the median [Fe/H] shows a slight decrease with distance along the minor axis (Y) up toY ≃ 20′, but the metallicity gradient completely disappears beyond this limit. Also, in some fields a very metal-rich ([Fe/H] ≥-0.2) component is clearly present.Whereas the fraction of metal-poor stars seems to be approximately constant (within few percent) in all fields, the fraction ofmetal-rich and, especially, very-metal-rich stars varieswith position and seems to be more prominent in those fields superposedon the disk and/or with the presence of streams or substructures (e.g. Ibataet al., 2001). This might indicate and possibly traceinteraction effects with some companion, e.g. M32.

Key words. individual Messier number: M31, M32 - stellar populations -stellar photometry -

1. Introduction

As the nearest bright spiral and the most prominent member ofthe Local Group, M31 has played a central role in our evolv-ing understanding of stellar populations. Most notably, Baade’s(1944) identification of Population II in M31 initiated a true

Send offprint requests to: M. Bellazzini, e-mail:[email protected]

⋆ Based on observations made with the NASA/ESA Hubble SpaceTelescope, obtained at the Space Telescope Science Institute, whichis operated by the Association of Universities for Researchin As-tronomy, Inc., under NASA contract NAS 5-2655. These observationsare associated with proposal GO-6671.

paradigm shift which in large part remains valid to the presentday.

Therefore, studying in detail the stellar populations in M31and comparing their properties to those of their Galactic coun-terparts is obviously very important to understand the forma-tion and evolution history of these two galaxies. Our presentknowledge already points to some interesting differences, sup-ported by observational evidence (see Harris & Harris, 2002;Ferguson et al., 2002, and references therein), e.g.:

i) The globular cluster population in M31 is∼ 3 times largerthan in the Milky Way, and on average more metal rich (Mould& Kristian, 1986; Durrell et al., 1994, 2001; Barmby et al.,2000; Perrett et al. , 2002);

http://arxiv.org/abs/astro-ph/0212531v1

2 Bellazzini et al.: The Andromeda Project I

ii) Also the abundance of the field population, in those fewcases where it is derived from calibrated colors, seems gener-ally higher than in the Milky Way, even at large galactocentricdistances where the halo population dominates (cf. Durrelletal., 2001; Ferguson et al., 2002; Rich et al., 1996, and refer-ences therein). A similar result was found for the halo starsinNGC 5128, with a striking difference compared to the MilkyWay (Harris & Harris, 2000, 2002). This has an impact ongalaxy formation models, in particular those based on accre-tion of disrupted satellites, since dwarf galaxies are metal poor(Da Costa et al., 2002, and references therein).

iii) Pritchet & van den Bergh (1994) found that “a single deVaucouleurs law luminosity profile can fit the spheroid of M31from the inner bulge all the way out to the halo”. However, clearevidence of “disturbances” such as spatial density and metallic-ity variations has been found, that can be interpreted as streamsand remnants of tidal interactions (Ibata et al., 2001; Fergusonet al., 2002).

The available data, therefore, would point toward interac-tion/merger events as non negligible factors in the building ofthe M31 halo. Therefore, very deep and detailed studies of thestellar populations and abundance distributions in the M31haloand outer disk are needed, as they can play a fundamental rolein explaining the formation of the M31 halo.

The advent of new technology detectors and telescopes, inparticular the Hubble Space Telescope (HS T), has boosted anew generation of studies, providing a much deeper insightin the understanding of the stellar content and evolutionaryhistory of this galaxy. In line with these recent efforts (e.g.,see Durrell et al., 1994; Morris et al., 1994; Couture et al.,1995; Rich et al., 1996; Holland et al., 1996; Holland, 1998;Guhathakurta et al., 2000; Ferguson & Johnson, 2001; Durrellet al., 2001; Stephens et al., 2001; Sarajedini & Van Duyne,2001; Ferguson et al., 2002; Williams, 2002, and referencestherein), we present here the first results of a large systematicsurvey of the stellar population in the disk and halo of M31 per-formed with the Wide Field and Planetary Camera-2 (WFPC2)on boardHS T.

The HS T program GO-6671 (PI R. M. Rich) was aimedprimarily at imaging with the WFPC2 a sample of bright glob-ular clusters in M31 spanning a range in metallicity. The glob-ular cluster target is placed on the Planetary Camera (PC) field,while the adjacent halo or disk field population of M31 fallson the 3 Wide Field Camera (WFC) chips. Adding to these7 more fields adjacent to globular clusters, whose images aretaken from theHS T archive, brings a total of 16 deep fieldsimaged byHS T, over a distance of about 4.5 to 35 kpc fromthe nucleus. More M31 fields are available in theHS T-archive,but we analyze here only those obtained in a strictly homoge-neous way (i.e. same camera and filters, similar exposures).

Photometry of the clusters is considered in a separate paper(Rich et al., 2003). In this paper we consider only photometryof the fields imaged by the WFC (Wide Field Camera) chips.Our photometry generally reaches to∼ 1 mag fainter than thehorizontal branch.

Our primary aim is to derive the abundance distribution ofthe M31 fields assuming that the population is old and globularcluster-like. This assumption is well justified by deep photom-

etry (e.g. Rich et al., 1996) that shows the subgiant luminosityfunction to approximately match that of the old globular clus-ter 47 Tuc. The abundance distribution of the M31 field stellarpopulation is then derived by interpolating between empiricaltemplate globular cluster RGB ridge lines, a technique firstap-plied by Mould & Kristian (1986) and later used in most of thesubsequent studies, though with significant differences.

The paper is organized as follows. Section 2 discusses theproperties of our field locations within the Andromeda galaxy.Section 3 discusses our observations and data analysis. Section4 presents the individual color-magnitude diagrams, includingan analysis of the young main sequence (blue plume) stellarpopulation. Section 5 describes our method for deriving theabundance distribution of the old stellar population. Section6 reports the abundance distributions, while Section 7 consid-ers the error budget (sensitivity to parameters such as redden-ing, distance modulus and abundance scale). We compare ourabundance distributions with those in the literature in Section8, while Section 9 considers how the detailed abundance dis-tributions vary with position and environment. We summarizeour results in Section 10.

2. The sample

M31 is a huge object in the sky, its optical angular diame-ter as reported by Hodge (1992) is 240arcmin, the absolutemagnitude∼ MV = −21.0. The galaxy has an inclination an-gle of i = 12.5 deg (Simien et al., 1978; Pritchet & van denBergh, 1994), and the apparent axis ratio is∼ 0.65 (Walterbos& Kennicutt, 1988; de Vaucouleurs, 1958). Three main stellarstructures have been identified: a wide exponential disk withspiral structures, a bright central bulge and an extended halo.

The peculiar inclination has prevented any firm conclusionon the presence of a thick disk as the one observed in our ownGalaxy (but see Sarajedini & Van Duyne, 2001, for a differ-ent view). Following Pritchet & van den Bergh (1994) we willconsider the bulge and the halo as a single component, thespheroid. We stress that this choice is made for the sake of sim-plicity and doesn’t imply any prejudice about the formationofthe halo and the bulge or the relation between them.

In Fig. 1, the positions of the observed fields are overplot-ted on a∼ 1 deg×1 deg image of M31, obtained using theTautenburg Schmidt telescope. Our fields sample the disk andhalo of M31 over a wide range of radii and positions, from∼20 arcmin (∼4.5 kpc) as far as∼ 150 arcmin (∼ 35 kpc) alongthe major and minor axes.

Table 1 gives the observed fields, which are designatedby their associated globular cluster (using both Sargent etal.(1977) and Battistini et al. (1987) designations). The rectan-gular coordinates X and Y [arcmin] are reported in column 3and 4, where the X axis coincides with the major axis of M31,and the Y axis coincides with the minor axis.R =

√X2 + Y2,

the radial distance from the center of the galaxy, is reportedin column 5, and the adopted reddeningE(B − V) is reportedin column 6 (see sect. 2.2). Finally, we report in column 7 theapproximate fraction of spheroid stars over the total (FS ph), asestimated in sect. 2.1 (but see also Sect. 9,10 for discussion).

Bellazzini et al.: The Andromeda Project I 3

2.1. Expected contributions to the observed samples

Only the outermost lines of sight in M31 sample a pure halopopulation, and clear traces of a disk population are found evenat those locations somewhat outside the faint disk isophotes.Therefore, we image some combination of disk and halo inmost fields.

It is possible to derive an approximate estimate of the rela-tive contribution of the disk and the spheroid along a particulardirection using the models by Walterbos & Kennicutt (1988)which reproduce very well the surface brightness profiles ofboth components over a very wide range (see also Pritchet &van den Bergh, 1994). The fraction of light must scale as thefraction of light contributors (stars), at least to first order. Sincethe Walterbos & Kennicutt (1988) models refer to the majoraxis and minor axis profiles, reasonably good estimates can beobtained for the fields that lie in the vicinity of the axes.

The fields G33, G87, G287 and G322 can be consideredto lie on the major axis. All of these fields are largely domi-nated by disk population. However, while G33, G322 and G287are sufficiently far from the center and suffer only a marginalcontamination by halo stars (∼ 7% for the first two fields and∼ 12 % for the latter), the growing importance of the bulge hassome influence on the G87 field (∼ 18% of contribution by thespheroid).

The fields G76, G119 and G272 are clearly projected ontoprominent features of the disk and their distance from the majoraxis is< 12′. Given their position with respect to the bulge, thecontamination by spheroid stars can be assumed to be< 10%.Obviously, the model of Walterbos & Kennicutt (1988) doesn’tinclude disk substructures such as spiral arms, and some fluctu-ation is possible. In particular the line of sight of G76 samplestheR= 40′ring, the site of the most vigorous star formation inM31 (Hodge, 1992; Williams, 2002).

The fields G108, G64 and G58 are relatively close to theNW arm of the minor axis, so the contribution of the variouscomponents to their population mix can be more correctely es-timated from the minor axis profile. The relative contributionsof the spheroid are∼ 60 %,∼ 65 % and∼ 70 %, respectively.These lines of sight intersect the halo, but the contribution fromthe disk population isvery important.

For the other fields a clearcut estimate based on the axis-oriented profiles cannot be done. However, all of them are far-ther than one degree from the center of M31 and most are ex-pected to be dominated by the halo. In particular G319, notvery far from the minor axis and atY = −68.9 arcmin, can beconsidered a “pure halo” field; G11, G219 and G351 all haveY > 40′, and the contamination by disk stars is expected tobe low (< 10 %). G105 hasY = −29.8′ and, judging fromthe minor axis profile the minimum contribution from the diskpopulation should be∼ 20 %. Finally, G327 hasY = 5.1′ but ismore than 2 deg from the center of the galaxy. The halo contri-bution as formally estimated from the major axis profile (withconsiderable uncertainty) is∼ 6 %, but the large distance fromthe center of the galaxy suggests that it may be considerablylarger than this.

Contamination by foreground stars belonging to our ownGalaxy and by background distant galaxies and quasars is also

possible. This issue has been extensively discussed by Hollandet al. (1996) who found the contamination from such sourceson the final CMDs and luminosity functions to be negligible.We repeated their tests and confirm their results.

3. Observations and Data Reduction

The observational material is described in detail in Table 2,which gives: (1) the field identification, (2,3) right ascensionand declination of the associated PC pointing, (4) date of theobservations, (5,6) identification number of theHS T programthat obtained the observations, and name of the program PI,and (7) filter and exposure time of the available CCD frames.

Most of the images are from the WF chips of exposures inwhich the PC was centered on the globular cluster (which wasthe prime target). One globular cluster, G327, was missed; ourfield is 36′ from the cluster but we preserve our usual namingconvention.

The G272 and G351 fields are parallel WFPC2 imagesfrom GO-5420, which imaged G280 and G351 using the FOC.Both fields are located at< 5′ of the respective clusters.

The G287 field was accidentally observed twice: G287b isslightly rotated relative to G287a. These data enable us to testour photometry (Sec. 3.4). Although we consider 17 fields, oursurvey actually has only 16 independent lines of sight.

The principal characteristic we want to stress here is thegreat homogeneity of the dataset. Of the 17 fields, 11 were im-aged as part of the same proposal, and the integration timesin each filter are identical. The other six image sets have beenselected because of similar exposure times and choice of fil-ters. Furthermore, all of the repeated exposures in each setweretaken at fixed pointing and we verified that the relative positionof each image within a data set coincides with the others towithin a small fraction of a pixel. This allowed us to producestacked images without any shift and/or interpolation (see be-low).

The distribution of stars within each field is very ho-mogeneous over the scales sampled by the WFPC2 camera.Image crowding is an issue in all cases, with significant vari-ations between the different fields, ranging from 640 resolvedsources/sq. arcmin in the outermost halo fields, to 13000 re-solved sources/sq. arcmin for those fields nearest the nucleus.Obtaining good photometry in such conditions is very difficult,since even the best PSF-fitting packages are subject to signifi-cant errors both in the photometry and in the interpretationofthe detected sources. This class of problems are the most sub-tle and dangerous because they permit contamination by largenumbers of spurious sources (e.g. see Jablonka et al., 1999).Since most of our fields are not lacking in sample size, we im-posed quite severe selection criteria in order to clean our sam-ple from spurious or badly measured sources. In some cases,up to 20% of the detected sources are rejected, under well de-fined and fully reproducible criteria which are described inthefollowing subsections.

After several tests that allowed us to optimize all the pa-rameters involved in any step of the data reduction process,weassembled a fully automated pipeline where the only input data


are an estimate of the background level of the images, the ex-posure times and the dates of the observations. The pipelinecarries out all the steps that are described in detail in the nextsubsections, from the PSF-fitting photometry on the stackedand cosmic ray cleaned images, to the production of the finalcalibrated and selected catalogue, every step being carefullychecked.

The only key step requiring manual intervention is the de-termination of aperture corrections, where obvious bad mea-sures or erroneous associations must be removed by hand (seesect. 3.2). Thus, also the data reduction process is strictly ho-mogeneous for every image set. Its reproducibility has bencarefully tested by comparing the results obtained by differentpeople reducing the same image set, and by the comparison ofthe results of the independent reduction of the two overlappingfields G287a and G287b. We stress the homogeneity of the dataset and of the data analysis procedure because we believe thatthis is a fundamental issue for a systematic study, as the onepresented in this paper.

3.1. Relative photometry

All the analysis was performed on the bias subtracted and flatfield corrected frames from the STScI pipeline. The overscanregion was trimmed and each set of repeated exposure imageswas coadded and simultaneously cleaned by cosmic ray spikeswith the standard utilities in the IRAF/STSDAS package.

The PSF-fitting photometry was performed using theDoPHOT package (Schechter et al., 1993), running on twoCompaq/Alpha stations at the Bologna Observatory. Weadopted a version of the code with spatially variable PSF andmodified by P. Montegriffo to read real images. A quadraticpolynomial has been adopted to model the spatial variationsofthe PSF. The parameters controlling the PSF shape were set inthe same way as Olsen et al. (1998), that successfully appliedDoPhot to the analysis of WFPC2 images.

Inspection of the F814W images shows that they have ahigher S/N in general than F555W despite their being exposedfor equal times in most cases. This is not surprising given thatmost of the detected stars are old, metal rich RGB stars. A fewpreliminary tests convinced us to use the F814W frames for thesource catalog, classifying as valid those sources brighter than3 times the local background noise. We then forced the code tofit the same sources in the F555W images, using the so called“warmstart” option of DoPHOT.

Since DoPHOT provides a classification of the sources, weretained only the sources classified as stars (type 1, 3 and 7). Wecross-correlated the F555W and F814W output catalogues witha tolerance of 1 pixel and finally produced a catalogue contain-ing, for every image set, the positions in the frame, F555W andF814W instrumental magnitudes and the associated errors, aswell as a parameter connected with the shape of the sources(ext= extendedness parameter, see Schechter et al., 1993). Wefound that in all cases the bulk of the sources classified as starsare confined in the range−20.0 ≤ ext≤ 20.0, and consequentlywe excluded from the catalogue all the outliers.

The CTE corrections have been applied according toWhitmore et al. (1999).

3.2. Calibrations

To transform the instrumental magnitude from the PSF-fittingphotometry to the Johnson-Cousin and/or STMAG system it isnecessary to apply the aperture corrections to a radius of 0.5arcsec, i.e 5 WF pixels (Holtzman et al., 1995).

Aperture photometry at a radius of 5 px was obtained for allthe stars whose peak intensity was higher than ten times the lo-cal background noise and having no detected companion withina 15 px radius, using the Sextractor package (Bertin & Arnouts,1996). Average aperture corrections were obtained for eachfield using these stars. Deriving reliable aperture correctionsin such extreme (and uniform) crowding conditions is very dif-ficult [see, for instance, Jablonka et al. (1999)]. Furthermore,because of the heavy undersampling of the PSF in the WF cam-eras, the flux sampled in the 5px aperture is strongly dependenton the exact position of the peak of the PSF within the cen-tral pixel of the star image (see Biretta, 1996). Thus there isno unique correction for all the stars but one has to rely on anaverage correction.

The total uncertainty in the absolute photometry introducedby the aperture corrections is in general±0.05 mag, but canreach±0.11 mag in the worst cases, a non negligible uncer-tainty indeed. However, these are the intrinsic and unavoidablelimits of the WFPC2 as a photometer, at least in the presence ofheavy,uniformcrowding. It is interesting to observe that all theindependently derived corrections are very similar (i.e. stan-dard deviations are rather small), a fact that testifies the sta-bility of the whole aperture corrections procedure. No trend ofaperture corrections with position in the frame was detected towithin the quoted uncertainties.

Once the aperture corrections were applied, the instrumen-tal magnitudes have been reported to the Johnson-Cousin andSTMAG systems using the standard calibration relations pro-vided by Holtzman et al. (1995).

3.3. Cleaning from contamination

In Fig. 2 the errors in the relative photometry for all the mea-sured sources are shown as a function of magnitude. It can bereadily appreciated that the mean uncertainties and the limitingmagnitudes are determined primarily by the crowding condi-tions, and differences in exposure time are much less important.For all cases errors are smaller than 0.1 mag forV < 24.5. Acheck of the reliability of the errors provided by DoPHOT willbe described in the next subsection.

All the stars with an associated error (either in V or in Imagnitude) larger than three times the average error at theirmagnitude level were excluded from the sample, as well as allthe stars havingerrV ≥ 0.6 mag and/or errI ≥ 0.6 mag. The av-erage error as a function of magnitude was calculated applyinga 2−σ clipping average algorithm on 0.5 mag wide boxes. Thelines superimposed on the plots of Fig. 2 represent the 3× errV

or 3×errI thresholds actually adopted. Errors much larger than


the mean can originate from a number of reasons: bad interpre-tation by the PSF-fitting algorithm, exceptionally high and/orvariable background, proximity to a very bright and/or heavilysaturated source, proximity to chip defects, partial saturation,etc. In any case, our aim is to prevent contamination by spuri-ous sources and this selection criteria is a very effective one.

A careful comparison between the derived catalogues andthe corresponding images showed us that there were still threecategories of undesirable sources that passed all the selectioncriteria applied, i.e.:

1. Spurious stars detected in diffraction spikes and coronaearound heavily saturated stars (type a).

2. Faint spurious stars detected in luminous extended back-ground galaxies (type b).

3. Faint background galaxies misinterpreted by the code asstars (type c).

Identification and hand removal of the type a and b sourceswas completed for all the fields withnsource≤ 35000 stars (seeFig. 2). In the same fields some type c sources have been in-dentified by accurate visual inspection of the frames, but somefaint galaxies probably remain in the final catalogues. For theremaining fields (nsource≥ 48000) the crowding prevented anyfurther cleaning. However the spurious sources represent aneg-ligible fraction of the samples: 3545 sources of type a,b andcwere removed over a total of more than 100000 in the inspectedfields (< 3.5 %).

Fig. 3 shows the rejected sources superposed on the CMDof the G64 field. Most of them are type a sources that lieupon the real stars distribution, mainly populating the region(V − I ) ≤ 1.5, while sources of type b and c are redder thanthis value. The blue region of the diagram is little affected bythis kind of contamination. If the faintest objects are excluded(those withI ≥ 26), only 10 % of the spurious sources populatethe (V−I ) < 0.8, 65 % are found in the range 0.8 ≤ (V−I ) ≥ 1.5and 25 % are redder than (V − I ) = 1.5. Most of the spurioussources are rather faint, 58 % of them lie below the red HBclump (I < 24.5), while the upper RGB is nearly free of con-tamination, only 8 % of spurious sources being brighter thanI = 23.5.

We can safely conclude that any spurious sources possiblysurviving these cuts could not affect the interpretation of ourCMDs, that are largely dominated by bona fide stars; in par-ticular, the RGB and red HB that we consider in this study areclean.

3.4. Reproducibility of measures and reliability oferrors: a direct test.

The large overlapping area of the G287a and G287b fieldsmade it possible to perform a direct test of the reliability andreproducibility of our measures. As this is the most crowdedfield of the whole set, it assures that we are testing the entireprocedure under the most unfavorable conditions. Thus the un-certainties derived from this comparison are conservative.

Both images were independently reduced, from the rawframes to the calibrated and selected catalogue, includingaper-ture corrections. The comparison was finally performed for

each WF chip separately. First, it has to be noted that virtuallyall the stars in the overlapping areas were measured in bothimage sets and independently survived the selections. Thusthedetection and selection criteria are very stable and reliable.

In the upper two panels of Fig. 4 the magnitude differencesas a function of the G287a magnitudes are shown, for all thestars in common. Note that in 4 out of 6 cases, the absoluteaverage differences are< 0.02 mag and in all cases they are< 0.05 mag. This demonstrates that the uncertainties in thefield-to-field relative photometry are fairly small in the finalcalibrated catalogue, i.e. the whole sample of 17 data sets istied to a common photometric system that is homogeneous to< 0.05 mag. The reproducibility of the results is excellent.

Individual uncertainties in the relative photometry arenearly as important as the measurements themselves. DoPHOT,as well as many other popular photometry packages, derivesthe errors in each measurement mainly from the signal to noiseratio of the fitted images and from the errors in the fitting pro-cedure. While the approach is formally correct and usually pro-duces a correct ranking of the quality of the measures, it is notguaranteed that the absolute estimate of the error is correct.The ideal estimate should be derived from the dispersion ofmany repeated measures, a procedure clearly not viable in thepresent case. In the lower panels of Fig. 4 the differences be-tween the errors from DoPHOT (ǫVa, ǫIa) and the error on themean from the two measures obtained for the stars in common(σ<V>, σ<I>) are averaged over 1 mag intervals and plottedversus V and I magnitudes. The two quantities are well cor-related and their difference is generally lower than 0.03 mag.Furthermore, DoPHOT seems to slightly overestimate errors,thus the most dangerous occurrence is prevented, i.e. drawingconclusions that are not supported by thetrue intrinsic accu-racy of the data.

4. The Color Magnitude Diagrams

Final CMDs for all of the 16 fields are shown in Figures 5 and6. The diagrams are ordered (from top to bottom and from leftto right) according to the projected distance of the field fromthe major axis X (see Fig.1). The only exception is the fieldG327 that is the outermost one (130 arcmin from the centerof the galaxy) but is relatively close to the major axis and hasbeen plotted as last. Note that nearly the same order is obtainedranking the CMDs by increasing contribution from spheroidstars, as reported in column 7 of Table 1.

The limiting magnitude depends mainly on crowding. Thedeepest photometry reachesI ∼ 27 (G351) but on average isabout I ∼ 26. Stars brighter thanI ∼ 20.5 are partially sat-urated. Despite the large differences in positions among thefields, the diagrams are quite similar if one takes into accountthe differences in the number of sampled stars.

4.1. The main morphological features

Our color-magnitude diagrams display the following commonfeatures:


– A red and broad RGB, ranging from (V − I ) ≈ 1 to(V − I ) ≈ 4, with the RGB tip nearI ≈ 21, and a promi-nentred clump(RC) on the RGB and HB, at (V − I ) ≈ 1andI ≈ 24.3. In all fields the upper part of the RGB bendstowards redder colors and fainter magnitudes in theV vs(V − I ) CMD, a characteristic seen in metal rich globularclusters and caused by TiO blanketing in metal rich firstascent stars. These characteristics are consistent with a rel-atively high metallicity and a wide abundance range.The presence of the dominant red clump inall of the halofields isprima facieevidence for an old metal-rich stellarpopulation in the halo, which is most likely a mixture ofintermediate age RGB stars over the full metallicity rangeand metal-rich old halo HB and RGB stars. The high metal-licity is also consistent with the slope of the RGB.

– A blue plumeat (V − I ) < 0.5, reaching in some casesI ∼ 20, is very evident in some of the CMDs. This featurecan be identified with the main sequence of an intermedi-ate/young population (YMS).Despite the large number of sampled stars, the YMS isbarely noticeable in the G287 and G87 fields, while it isclearly present in G33, G322, G119, G272, G108 and G64,and it is strongly present in the G76 field.The existence of a sparse YMS cannot be excluded in thefields G58, G11 and G351. In the last two cases the low to-tal number of stars in the sample prevents any firm conclu-sion in this sense. In any case, the good anti-correlation be-tween the prominence of the YMS and the estimatedFS ph

is remarkable.– A small number ofblue HB starsat (V − I ) ∼ 0.5 and

I ∼ 24.5 is present in many of the halo fields (e.g. G11,G105, G219, G327, G319, G319 and G351), and could bealso present, but hidden within the YMS population, in theother fields (see e.g. G64).These stars indicate the presence of some globular-clusterage metal poor stars, because presumably only stars olderthan 12 Gyr can reach the blue HB. Their numbers (≈ 15%of the RHB) are consistent with the fraction of metal poorstars in the abundance distributions that we derive from theRGB.

– In the CMDs where the YMS is most conspicuous, aredplume is seen betweenI ∼ 22 andI ∼ 20, around (V −I ) ∼ 1.5. Because of the strong young stellar population,we interpret stars in this region as being red supergiants,the evolved counterpart of the YMS stars.

We represent the CMD of the G76 field as a contour plot(Hess diagram) in Fig. 7, in order to illustrate more clearlytheprincipal features present in all of our fields.

The most prominent feature common to all of the CMDs(other than the red giant branch) is the Red Clump, whichshows an elongated structure with a very sharp peak. The fewstars brighter thanI ∼ 21, beyond the RGB tip, could eitherbe cool RSG and/or bright AGB stars, the latter not necessar-ily indicative of the presence of intermediate age populations,[see Guarnieri et al. (1997)]. There is also a real possibility thatsome of these are simple blends of normal RGB stars, sincethis is expected in extremely crowded fields [see Jablonka etal.

(1999), in particular their figs. 5a and 5d, and see also Renzini(1998)].

An accurate treatment of the blends, requiring a large num-ber of artificial star experiments, is important for some issues(e.g. to constrain the fraction of intermediate-age stars in thehalo) that are beyond the scope of the present paper.

Figure 8a shows stars from the halo fields G105,G219,G319 and G327 are plotted together in the same (V,V-I) CMD,centered on the Horizontal Branch. To limit the contaminationfrom spurious measures, we plot only the stars with photomet-ric errors less than 0.15 mag in both passbands.

The BHB is clearly visible atV ∼ 25.2, extending from thered HB clump to (V − I ) ∼ 0, as previously found by Hollandet al. (1996). A schematic ridge line derived from the HB ofthe metal poor globular cluster M68 [data from Walker (1994)]is superimposed on the plot, applying a shift of 9.44 mag in Vand 0.09 mag shift in color.

Fig. 8b shows the CMDs of the fields G11 and G351, plot-ted as in panel (a). A BHB similar to that shown in panel (a)is still present, but in this case the feature is partially contam-inated by fainter stars looking like the base of the YMS blueplume described above, and this is the reason why the CMDsare plotted separately. Considering the distance of G11 andG351 from the disk, the blue stars might also be field blue strag-glers.

As G11 and G351 are among our deepest fields, we cannotexclude such a blue plume from the other fields either. Giventhe complexity of the M31 halo, it is possible that intermediate-age populations are distributed throughout the halo in varioustidal streamers, present in some fields and not (or less) in oth-ers. The only secure determination of the nature of these faintblue stars will be imaging deep enough to reach the old mainsequence turnoff at MV = +5. However, even from the presentdata sets we can already deduce that subtle and possibly impor-tant differences may exist in the stellar populations of the outerfields of the M31 halo.

4.2. The intermediate-young population

In those fields with a substantial disk contribution, the presenceof a blue plume is evidence of star formation within the last 0.5Gyr. The power of our survey is that we can use deep imagingover a range of fields, enabling us to probe well past the fewMyr of star formation history revealed by the brightest OB starsand HII regions.

Any attempt to derive the star formation history (SFH) fromthe CMDs and LFs would need a detailed analysis based onthe comparison with appropriate synthetic evolutionary tracksand taking into account all the observational effects [see, e.g.Tosi et al. (1989), and Aloisi et al. (1999); Gallart et al. (1999);Hernandez et al. (1999) as examples of recent applications].Williams (2002) performed a similar type of analysis in M31using 27 fields imaged byHS T along the disk, in order to de-termine the star formation history that best fits his observations.

This type of detailed study is beyond the scope of thepresent paper; however, it is possible to apply a simple anal-ysis to our data and obtain useful information to constrain


the relevant timescales and stellar masses associated withtheprominent YMS of many disk-dominated fields. This aim canbe easily achieved by comparison with theoretical isochrones,and would also help to set a basic interpretative scheme for theCMDs we present.

In order to compare our CMDs to theoretical quantities,however, we need to make assumptions on two basic param-eters, reddening and distance modulus.

– Reddening− Interstellar extintion toward M31 can be dueto dust screens residing either in our Galaxy or internal toM31 itself. While a reliable estimate of the average redden-ing due to the Milky Way can be obtained from reddeningmaps (Burstein & Heiles, 1982), estimates of the “intrin-sic” reddening are not available and this is one of the majorsources of uncertainty in the study of stellar populations inM31 [see the thorough discussion by Barmby et al. (2000)].These latter authors attempted to estimate the total redden-ing toward the M31 globulars, calibrating the relations be-tween integrated colors (essentially B-V) and metallicitywith Galactic globulars for which the amount of reddeningis known from independent measures.We have applied a similar relationship to the integratedcolors and metallicities of the globular clusters associatedwith the fields that are projected onto clear disk structures,i.e. that are more likely affected by intrinsic extinction.For G76, G87, G119, G272 and G287 we find an average< E(B − V) >= 0.14 andσE(B−V) = 0.03. We decided toadopt this value since the single values were remarkablysimilar and to avoid the introduction of noise due to theuncertainties of the calibration.For all the remaining fields that are distant from the highsurface brightness regions of the disk we adopt the Galacticreddening toward M31. Following Hodge (1992), we adoptE(B − V) = 0.11± 0.02 from McClure & Racine (1969);we use the Dean, Warren & Cousins (1978) extinction lawE(V − I ) = 1.34E(B− V) andAI = 1.31E(V − I ).

– Distance−We adopt the Cepheid distance modulus of (m−M)0 = 24.43±0.06 (∼ 770 kpc) from Freedman & Madore(1990). Other distance estimates may be found in Table 3.1in van den Bergh (2000).

4.2.1. The age range of the YMS stars

Fig. 9 shows the CMD (inMI , (V − I )0 for the sum of the disk-dominated fields G76, G119, and G322 (154,748 stars). Thestars above and below theMI = −2 threshold are plotted usingpoints of different thickness to allow easier recognition of boththe densely populated features in the lower part of the CMD(e.g. the HB clump) and the sparse bright features (e.g the up-per MS and the red plume of RSG stars).

Six isochrones with solar metallicity and helium abundanceY = 0.28, from the set by Bertelli et al. (1994), are superim-posed to the diagrams. They correspond to ages of 60, 100,200 and 400 Myr (from top to bottom, continuous lines), 1 Gyr(open squares) and 12 Gyr (open circles). Our adopted compo-sition seems the most appropriate for intermediate/young pop-ulations in the disk of a large spiral galaxy.

Before we consider Fig. 9, recall that the brightest MS starsknown in M31 reachMV ∼ −6.2, but the saturation level in oursurvey occurs atMV ∼ −4, which eliminates the brightest mainsequence stars from our sample.

The isochrones of 60, 100, 200 and 400 Myr fit well theobserved distribution of young and intermediate age stars,thusconstraining the age range associated with the sampled YMS.

The RSG branches of these isochrones group between (V−I )0 ∼ 1 and (V − I )0 ∼ 1.6, providing an excellent fit to the redplume described in the previous section (figures 6 and 9). The1 Gyr isochrone illustrates clearly how intermediate age AGBstars can populate the region of the CMD immediately abovethe RGB Tip.

As a final consideration we note that RGB stars as red as(or redder than) the 12 Gyr isochrone are clearly present in thecomposite CMD of Fig. 9. The obvious conclusion is that veryold and very metal rich stars are found in the disk of M31.

4.2.2. A Classification Scheme for the Blue PlumePopulation

In order to model the main sequence luminosity function wehave grouped stars by luminosity bins so that they crudely sam-ple different mass ranges. We can use these counts to explorethe relative importance of the YMS in different fields.

In Fig. 10 we overplot 3 boxes on the (V,V-I) CMDs ofthe G64 field, as an example. These boxes are rather large andclearly separated in order to include a portion of a particularsequence independent of any slight difference in reddening ormetallicity among the different fields.

The regions are defined as follows:

1. Yu: (23.5 < V ≤ 22.5 and−0.3 ≤ V − I ≤ 0.6) samples theupper YMS. The number of stars in this part of the CMDsare indicated asNYu.

2. Yd: (24.5 < V ≤ 23.5 and−0.3 ≤ V − I ≤ 0.6) samples thelower YMS. The number of stars in this box are indicatedasNYd.

3. C: (24.5 < V ≤ 23.5 and 0.9 ≤ V − I ≤ 2.6) samplesthe RGB in the same magnitude range as Yd. The numberof stars in this box are indicated asNC. We will useNC tonormalize the star counts in the previous boxes.

According to the same set of isochrones shown in Fig. 9 itcan be stated that theYd box samples stars younger than∼ 0.5Gyr while theYu boundary samples stars younger than∼ 250Myr. We thus derive the following indices::

– (NYu + NYd)/NC: ranks the fields according to the relativeimportance of the YMS population.

– (NYu/NYd): since more massive (i.e. younger) YMS starsare expected to reach brighter magnitudes, this ratio testsdifferences in the recent star formation history between thedifferent fields.

The errors in the star counts have been estimated accord-ing to Poisson statistics, and we propagate the errors to de-rive errors in the indices. It is obvious that our detection of the


bright MS is most significant for the most populous, disk dom-inated fields. The fields G11, G58, G105, G219, G319, G327and G351 have too few young stars to be worthy of furtherconsideration in this sense. As already noted, this doesn’tnec-essarily mean that no young stars are present in these fields,since the low total stellar density can prevent the detection ofshort lived stars. However, a comparison with the most simi-lar field having a significant YMS population, i.e. G64, showsthat these fields are intrinsically deficient of “Yu+Yd” stars bya factor∼ 2− 5 with respect to G64.

Fig. 11 shows (NYu + NYd)/NC (panel a) and (NYu/NYd)(panel b) plotted as a function of the distance from the centerof the galaxy, deprojected to “face-on”Rp[arcmin] (see Hodge,1992, and references therein). The deprojection has been ob-tained adoptingi = 12.5 deg, and it is justified in the presentcase since we are dealing with (i) mostly disk dominated fieldsin the plane of M31, and (ii ) the YMS population which is ex-pected to be located in the star forming disk.

The prominent feature appearing in panel (a) is the strongpeak in the relative abundance of stars younger than∼ 0.5 Gyroccurring atRp∼ 60′, shown mostly by G76 but also by G119and G322. This corresponds to theR ∼ 10 kpc ring in whichneutral hydrogen and virtually any tracer of a young popula-tion seem to cluster (Hodge, 1992; van den Bergh, 2000). Thismain structure of the star forming disk is clearly sampled byour fields as a very significant enhancement in the recent starformation rate: there is a factor∼ 35 in the relative abundanceof YMS stars between the nearly quiescent inner regions (fieldsG87, G287) and the peak of the ring structure (field G76).

In panel (c) of Fig. 11, we show the (NYu+ NYd)/NC indexas a function of the absolute distance along the major axis (|X|),to provide a different view of the results presented in panel (a).

In particular, this panel allows a direct comparison with thedistribution of disk tracers as depicted by Hodge (1979) (seehis Fig. 7 and 8). The agreement is indeed remarkable, and thegeneral picture assembled by Hodge (1992) is confirmed byour survey, at least to the extent permitted by our limited spatialsampling1.

Panel (b) shows that also the star formation in the last 250Myr was particularly strong near the position of the G76 field.However, the most noteworthy feature is the highNYu/NYd

value in the innermost field G87, indicating that most of the(weak) star formation in this field occurred quite recently.

This is particularly interesting in comparison with the oth-erwise similar field G287 that shows aNYu/NYd a factor∼ 5lower for the same value of (NYu+ NYd)/NC.

As a result of these considerations, we can conclude thatthe star formation rate has varied significantly over the last 0.5Gyr in different regions of the disk of M31. Although our find-ings are not based on as sophisticated an analysis as that ofWilliams (2002) we agree with the conclusions of that study.It is noteworthy that the observable activity in the 10 kpc ringcorresponds to recent significant star formation, and that recentstar formation activity is not necessarily correlated witha high

1 It is important to recall that the work by Hodge (1979) is based ona sample of 403 open clusters while we are presently analysing just 9small fields.

total rate of star formation. Surveys with theHS T− ACS andmore sophisticated modelling will greatly improve our under-standing of this issue.

5. Assumptions and procedures for metallicitydeterminations

It is well known that the color of the RGB of an old SimpleStellar Population [SSP, i.e. an ensemble of stars sharing thesame age and chemical composition, see Renzini & Buzzoni(1986)] is mainly affected by the abundance of heavy elementsand only at a lesser extent by the age of the population. Theinfluence of age becomes smaller and smaller with increasingage and it is almost negligible for ages in excess of∼ 10 Gyr. Inprinciple, the metallicity of a RGB star is uniquely determinedby its color and luminosity, once its age is known and a suitablecalibration is available. The obvious application of such prin-ciple is to derive the metallicity distribution for a given popu-lation from the distribution in color and magnitude of its RGBstars (see Saviane et al., 2000).

However, when dealing with a Composite Population (CP),i.e. a mix of stars of different ages and metallicities, there aretwo main factors that can affect the correct recovery of the un-derlying metallicity distribution:

– AGB sequences, in general, run nearly parallel to the RGBbut are slightly bluer. In the absence of any useful crite-rion to exclude them from the sample, they would appearas more metal deficient than their parent population, intro-ducing a bias in the metallicity distribution (see Holland etal., 1996).The existence of a blue HB (as well as the results of thespectroscopic survey by Guhathakurta et al., 2000) tells usthat a metal poor old population exists for sure. Hence, theblue side of the RGB must contain their metal poor giantprecursors. If this is the case, contamination from AGBstars is necessarily very small since their lifetimes are sig-nificantly shorter than lifetimes on the RGB. In fact, for asolar metallicity population of age 15 Gyr, the number ofRGB stars is predicted to overwhelm the number of AGBstars by a factor∼ 40 (Renzini, 1998). Based on the aboveconsiderations, only a very small fraction of our metal poorgiants may, in principle, be misclassified AGB stars.

– Age differencesat fixed metallicity produce a widening ofthe RGB that could be erroneously interpretated as a metal-licity range. This effect is rather subtle, since it depends onthe whole star formation and metal enrichment histories ofthe composite population under study.For example the RGB of an old population (say∼ 12Gyr) of solar metallicity can be contaminated by stars froma population of the same metal content but several Gyryounger that would simulate the presence of older metalpoor stars [see sec. 3.6.1 of van den Bergh (2000)].The metallicity distributions (MDs) derived from the colordistributions of RGBs in our CMDs are based on the as-sumption that the observed RGBs are dominated by oldstars.


This hypothesis is not at odds with observational evidence[see Tinsley & Spinrad (1971); van den Bergh (2000);Hodge (1992), and references therein], and all the numer-ous attempts made in previous studies are based on it (Richet al., 1996; Holland et al., 1996; Pritchet & van den Bergh,1994; Richer et al., 1990; Morris et al., 1994; Jablonka etal., 1999, 2000; Durrell et al., 1994; Couture et al., 1995;Durrell et al., 2001; Ferguson & Johnson, 2001; Harris &Harris, 2000, 2002; Sarajedini & Van Duyne, 2001).However, our previously stated caveats should be kept inmind when drawing any conclusion from the derived abun-dance distributions. This is especially true for the abun-dance distributions measured in those fields dominated bydisk stars.

In addition to the above issues related to a composite popu-lation, there is a third problem one must take into account, i.e.the differential reddening that might affect the photometric dataespecially in the disk dominated fields.

Differential reddening does not seem to affect significantlyany of our fields at a cursory inspection: for example, the wideRGB is found in distant halo fields, and no dispersion of the redclump along the reddening vector is seen in any field. Further,the blue main sequence, where present, is not similarly broad-ened.

However, we tested possible variations in interstel-lar/intergalactic extinction across the observed fields by com-paring the CMDs obtained from different subsections of eachsingle WFPC2 field. In every case we failed to detect any dif-ference in the location of the main branches, and we concludethat variations of the extinction across the observed fields, ifany, are negligible with respect to the intrinsic width of thebranches and the observational errors.

Variations of the extinction along the line of sight couldalso affect the CMDs, but this effect would be more subtle.Dust structures in M31 might be embedded in the population,with some stars in front and some behind the cloud. However,the striking similarity between MDs (see Sect. 6.2) of fieldssampling dense and active regions of the disk (f.i. G287, G33,G76) and those sampling outer halo lines of sight (f.i. G319,G351, G1), that are expected to be unaffected by M31 dust, ar-gues against this type of differential extinction. To obtain sucha result from a dataset where differential reddening has a sig-nificant effect in some fields and none in others would imply animplausible fine tuning between differential reddening, metal-licity distribution and star formation history.

We conclude that differential reddening has at most amarginal effect on our CMDs and MDs.

5.1. The grid of Galactic Globular Cluster Templates

In our analysis we shall adopt as a reference system the Carretta& Gratton (1997) (CG) abundance scale, that is tied to high-resolution spectroscopy and offers a better guarantee of accu-racy and reliability with respect to the Zinn & West (1984)(ZW) metallicity scale, that was based on photometric indices.However, the ZW scale has been and still is widely used, so

we shall comment our results by comparing the effects of bothscales (see Sect. 7.1).

For a detailed description of the CG vs. ZW scales we re-fer to Carretta & Gratton (1997). Here we just note that theCG scale yields more metal-rich values in the interval−2 <[Fe/H] < −1 (this effect disappearing progressively as onemoves towards the edges of this interval) and more metal-pooroutside. Also, the metal-rich extension towards solar values isstill poorly sampled and very uncertain for both scales, withCG yielding slightly more metal-poor values than ZW.

As we will discuss below, the choice of the CG or ZWmetallicity scale has little effect on our main conclusions con-cerning the overall properties of the derived MDs. In fact, thesame general conclusions can be drawn by a direct compar-ison of the star color distributions with the grid of the RGBridge lines for Galactic GCs of known metallicity. However,the detailedshape of our abundance distributions and, there-fore, somespecificconclusions may indeed be affected by thischoice (and, more significantly, by the adopted globular clustergrid and interpolating procedures).

Finally, we caution the reader thatsystematiceffects onthe zero-points of magnitudes and colors or of the metallic-ity scales may actually have the strongest impact on the overallpicture.

For the metallicity determinations we adopt an approachsimilar to Holland et al. (1996), i.e. we compare the observedRGBs with a grid of accurately chosen RGB fiducial ridge linesat various metallicities, and then derive a metallicity estimatefor each star by interpolating from the grid of templates. Weapply the technique in the [MI , (V − I )0] plane since in thisplane the behaviour of the RGB sequence is less susceptible of“curvature” effects (see Saviane et al., 2000).

As RGB templates we adopt the ridge lines of the galacticglobular clusters NGC 6341 ([Fe/H]CG = −2.16), NGC 6205([Fe/H]CG = −1.39), NGC 5904 ([Fe/H]CG = −1.11) and 47Tuc ([Fe/H]CG = −0.71) from Saviane et al. (2000).

As noted by all the authors adopting the present approach,the extension of the grid to more metal-rich values than 47Tuc is quite difficult (for the lack of suitable candidates) andextremely uncertain (for intrinsic uncertainties in the high-Zregime).

After a careful revision of the available data, we choose toadopt for the very metal-rich regime two clusters, NGC 6553and NGC 6528, for which sufficiently good V,I data and metal-licity estimates (for both CG and ZW scales) are available.

Therefore we completed the reference grid with the ridgeline of:

NGC 6553 ([Fe/H]CG = −0.16, Cohen et al. (1999)) ob-tained from the photometric data of Guarnieri et al. (1997),and

NGC 6528 ([Fe/H]CG = +0.07, Carretta et al. (2002)) ob-tained from the photometric data of Ortolani et al. (1995).

Reddenings and distance moduli are taken from the com-pilation by Ferraro et al. (1999), since their approach in themeasure of distance moduli is homogeneous for all the clustersand independent of their HB morphology or of the presenceof RR Lyrae variables. Table 3 includes these values and themetallicities of our calibrating clusters.


Before proceeding further, we wish to add a few considera-tions on the problems and implications the choice of a differentgrid may have on the subsequent analysis.

All previous studies dealing with a similar procedure ofderiving the MDs from the color distributions of the giantbranches in the CMDs have adopted their own custom-tailoredrecipe and reference grid (see Sect. 8). Some authors haveadopted purelyempirical grids or, alternatively, purelytheo-retical ones, others have adopted amixtureof both, using GCdata to set the zero-point of a given set of theoretical models,or adding suitably calibrated isochrones to empirical templatesto cover missing parts of the metallicity range. Furthermore,within each adopted procedure, the choice of selected GCsand/or set of models differs from study to study. Though thebulk of the main results is probably independent of these dif-ferences, there is no doubt that the detailed shapes and prop-erties of the MDs are strongly affected and, especially in thevery high-metallicity regime, these choices may dominate theresults.

The reader should be aware that the same original data,transformed into MDs with different recipes, may yield signif-icantly different resultsin some details. Coupled with the in-trinsic, quite high uncertainties still affecting the observationaldata-bases themselves, one has to admit that all our analysesare still at a rather preliminary stage and far from offering anunambiguous detailed comprehension of this issue (see Harris& Harris, 2002, for discussion).

We have emphasized above that our reference grid reliesonly on empirical templates, chosen to be a homogeneousand reliable set, and sufficiently sampled to cover the relevantmetallicity range in suitably fine steps. The accuracy of thisapproach depends only on observable quantities, i.e. photomet-ric and spectroscopic data and reddening estimates, whose er-rors can be known and minimized to some extent, within thepresent possibilities. The use of theoretical models, thatwouldbe in principle easier and more precise, suffers of its own setof problems. In fact, while great progress has been made inproducing synthetic RGB sequences, the physics of late-typestellar atmospheres (e.g. convection, alpha enhancement,etc.)and the color-temperature calibration, bolometric corrections,etc. are still poorly known, and the use of theoretical RGBsmight introduce additional uncertainties which are difficult tocorrectly estimate and quantify.

5.2. The procedure of metallicity determination

Both the M31 CMDs and the templates are transformed tothe absolute plane, using the values of reddening and distancequoted in Sect. 4.2 for M31, and listed in Table 4 for the tem-plates.

The estimates of metallicity are performed on stars having−3.9 < MI < −2.0 and 0.90 < (V − I )0 < 4.0. The reasons ofthis choice can be summarized as follows:

– Lower Luminosity limit:MI < −2.0to retain the region of the RGB with the highest sensitivityto metallicity variations and to avoid contamination by RCstars. This choice also avoids the inclusion in the sample of

the AGB clump stars, which are predicted (and observed) tolie around−1 < MI < −1.6 (see Ferraro et al., 1999). Thusthe more densely populated feature of the AGB is excludedand we can be confident that only a marginal fraction of oldAGB stars may contaminate our metallicity distributions.

– Upper Luminosity limit:MI > −3.9to avoid the inclusion of bright AGB stars. To examinein detail the possible impact of different bright cuts wehave carried out several tests (cutting at different luminos-ity thresholds and using different interpolating schemes; theone actually used is illustrated in Fig. 12 and 13). Thesetests have produced insignificant deformations in the MDs,and varying the upper luminosity cut has a negligible effecton the basic morphology. Therefore, in the following weshall use the above bright limit because it adds statisticalsignificance, especially in the poorly populated halo fields.

– Blue color limit : (V − I )0 > 0.90to limit the contamination by AGB stars and young stars(see Fig. 9).

– Red color limit: (V − I )0 < 4.0to avoid contamination by foreground and/or backgroundsources.

The interpolation procedure has been accurately checkedand tested, and proven to perform very well. We estimate thatthe uncertainty in a single metallicity measure is±0.2 dex (ran-dom error).

Our interpolation scheme is strictly self-consistent onlyover the metallicity range defined by our grid of GGC tem-plates, but we decided to allow a modest linear extrapolation tostars slightly bluer than the ridge line of NGC 6341 and slightlyredder than the ridge line of NGC 6528. The allowed extrapola-tion is of the order of the assumed uncertainty in the metallicityestimates, i.e. 0.2 dex: stars beyond these extrapolated limitsare excluded from the final metallicity distribution (MD). Thefraction of stars excluded from each sample is in general lessthan 2−3 % (see Table 4), and does not have significant effectson the description of the MDs.

6. Results: the Metallicity Distributions (MDs)

6.1. A first cursory inspection

We show in Fig. 12 the [MI , (V− I )0] CMDs of the upper RGBsfor four fields, taken as representative of the whole sample:G87 (the innermost one), G119 (a disk dominated field with arich YMS population), G64 (an intermediate field with signifi-cant contributions from both disk and spheroidal components),and G11 (a typical halo dominated field).

The template ridge lines are superimposed to each plot,from left to right: NGC 6341 (M92), NGC 6205 (M13), NGC5904 (M5), NGC 104 (47 Tuc), NGC 6553, NGC 6528. Theinner frame drawn on each plot shows the region of the CMDselected for the metallicity estimate, as described in the previ-ous section.

A cursory inspection of Fig. 12 shows the following quali-tative characteristics:


– The distribution of the stars with respect to the ridge linesis quite similar for the four cases, i.e. the bulk of them isbracketed by the ridge lines of 47 Tuc and NGC 6553, witha very wide spread. Taken at face value, this suggests thatthe metallicity distributions must peak somewhere in therange−0.7 ≤ [Fe/H] ≤ −0.4, independently of whichmetallicity scale is used for the detailed abundance esti-mates. This is in agreement with many previous results(Mould & Kristian, 1986; Richer et al., 1990; Durrell et al.,1994; Morris et al., 1994; Couture et al., 1995; Holland etal., 1996; Durrell et al., 2001; Ferguson et al., 2002; Harris& Harris, 2000, 2002).

– The width of the RGB distribution in M31 largely exceedsthe observational errors, which are less than 0.10 mag ineach bandpass. This result too has been found in all previ-ous M31 studies. Differential reddening, which could causesuch a spread, can be ruled out as it was discussed in Sect.5. Therefore, some metallicity (and possibly age) spreadmust be present.

– The statistical weight of the metallicity distributions variesfrom field to field depending on the density of the stellarpopulation: sparse halo fields can suffer from small numberfluctuations but are cleaner from contaminants, rich diskdominated fields have a larger number of stars but are moreprone to contamination by younger populations. Note how-ever that the adopted selection in color is quite effectivein preserving the samples from the pollution by undesiredstars in such fields (see the CMD of G76).

For a more quantitative and detailed analysis, we need toinvestigate the metallicity distributions in all our fields.

6.2. MDs of the individual Fields

Based on the ridge lines, the distance moduli and the redden-ings of the template clusters listed in Table 3, and on the as-sumptions we have made about the distance modulus and red-dening of M31 (Sect. 4.2), we have derived the metallicity dis-tributions for all the individual fields shown in figures 14 and15 in the same order as in figures 5 and 6.

In each panel is reported: the name of the field, the num-ber of stars used to derive the MD, the average metallic-ity ([Fe/H]ave) together with the associated standard devia-tion, the median metallicity together with the associated semi-interquartile interval, and the fractions of stars with [Fe/H] <−1.4 [Metal-Poor or Young – hereafter: MPorY],−1.4 <[Fe/H] < −0.2 [Metal-Rich – MR], and [Fe/H] > −0.2 [VeryMetal-Rich – VMR], respectively (see below).

The most striking property of the MDs shown in figures 14and 15 is their overall similarity. To describe them in a quan-titave way (without ”forcing” the analysis beyond a certainlevel of speculation), after several different tests we decidedthat the best solution would be to schematically subdivide thehistograms into three main intervals as indicated by the verticallines drawn in the plots presented in Fig. 14 and 15:

– [Fe/H] < −1.4, theMetal-Poor or Young group –MPorY– −1.4 < [Fe/H] < −0.2, theMetal-Rich group –MR

– [Fe/H] > −0.2, theVery Metal-Rich group –VMR

The values of –1.4 and –0.2 were chosen because they markeither a discontinuity in the MDs (at –1.4), or a discontinuousbehaviour of the metal-rich tail (at –0.2).

The long ”thin” tail reaching metallicities as low as[Fe/H] < −2.5 could be interpreted as due to: (a) very metal-poor old stars truly representative of the old halo of M31,though at a trace level, and/or (b) young blue stars which mimica metal poor population, especially in the inner disk fields.Thisis the reason why we have called this group MPorY.

On the opposite extreme of the distributions, in the VMR-regime, at [Fe/H] = −0.2, especially in some fields (e.g. G287,G87, G33, G322, G76, G219, G319), one can see a second dis-continuity (variable in size from field-to-field) which could beascribed to the existence of a very metal-rich component, addedto the metal-rich tail of the bulk population.

In a few fields (i.e. G108, G58, G351, G219, G327)one might also see another discontinuity at [Fe/H] ∼ −0.8.However, taking into account the quoted intrinsic uncertaintiesand the possible effects induced by the binning size, we are in-clined to adopt just two cuts. We will discuss in a specific sec-tion (sect. 9.3) the results of dividing the total sample in twosub-samples only, cutting at [Fe/H] ∼ −0.8.

Before summarizing it may be useful to note that we do notreport the results of any multi-gaussian fitting to the data,aswe found that the degree of discretionality in setting the severalparameters (number of components, metallicity limits, widths,etc.) is rather high compared to the intrinsic quality of theavail-able data.

In synthesis all the fields:

– have the bulk of the population located within the centralmetallicity interval, i.e.−1.4 < [Fe/H] < −0.2, with anobvious main peak around [Fe/H] ≃ −0.6 ± 0.20. Mostof the fields in Fig. 15 are so distant from the disk as tobe true halo fields, yet the metal rich peak is the most ev-ident feature of the distribution, and the fraction of starswith [Fe/H] < −0.8 ranges from only 0.1 to a maximumof 0.4 (see Table 5). To test further this point, we reducedthe frames used by Rich et al. (1996) and derived the cal-ibrated photometry of the field of the remote cluster G1(Rarcmin ∼ 150), with the same pipeline adopted for allfields presented here. The resulting CMD is virtually iden-tical to that presented by Rich et al. (1996). In Fig. 16 it isshown that the MD of this extreme region of M31 is verysimilar to that of the other inner fields. In particular, themean metallicity is still as high as [Fe/H] = −0.8.

– show a long, poorly populated tail spanning the metallicityinterval −2.5 < [Fe/H] < −1.4. As one can see in Fig.8, stars bluer (more metal-poor) than the ridge line of G11([Fe/H] ∼ -1.9) do exist and have been detected in all fieldsat the level of a few percent of the total sample. These starsare probably representative of the very metal-poor old haloof M31 in the outer, less disk contaminated fields. In theinner disk-dominated regions this subsample most probablyincludes blue young stars which in our procedure (based oncolors) may mimic metal-poor objects.


– show a significant population of stars with [Fe/H] > −0.2,varying with distance along the Y-axis from∼ 40% in themost central fields to less than∼ 4% in the most distantones, with a few exceptions (e.g. G219).

– the field of G219 stands out in that its VMR pupolationis stronger than might be expected from its location, andthe distribution looks qualitatively different from the others.As discussed later and shown in Fig. 20 , this field is wellsuperposed on the metal rich tidal tail found by Ibata etal. (2001) and Ferguson et al. (2002) (their Fig. 7). It isinteresting to note that the abundance range of this fieldis similar to those located at large galactocentric distance,only the prominence of the VMR peak is greater.A similar behaviour, albeit with a much wider MD and aless pronounced peak, might be present in G319 locatednot far from the quoted stream (see Fig. 20).

Table 4 reports the populations of each metallicity bin (step0.2 dex) for each field to make available to the reader the gridof the MDs for further analysis.

6.2.1. Is our Very Metal-Poor population real?

As mentioned earlier, metal poor stars are expected to bepresent, based on the existence of a blue HB. If the metal poortail were mostly due to contaminating younger stars it wouldbe expected to reach its minimum extent in the most exter-nal halo fields where no such stars are expected to be present.However, the fraction of blue objects seems to be fairly con-stant (∼ 2 − 5%), so we are inclined to interpret them as trulyold and very metal-poor stars.

Independent confirmation of the existence of metal-poorhalo giants has been provided by the recent work of Reitzel& Guhathakurta (2002) based on Keck/LRIS spectra. ¿Froma preliminary sample (obtained from a 16× 16 arcmin2 fieldlocated atR = 1.6 deg on the M31 minor axis), they selected24 bona fideM31 RGB stars on the basis of their position inthe CMD and their radial velocity, and derived metallicity es-timates from the strength of the Ca II lines. 21 of the selectedRGB stars have [Fe/H] ≤ −1, 17 of them have [Fe/H] ≤ −1.5and some reach metallicity of [Fe/H] ∼ −2.5 or lower. The re-sult is independent of the assumed metallicity scale. So metalpoor stars are indeed present in the M31 halo and are not rare.

6.2.2. Is our Very Metal-Rich population real?

As already noted by simple visual inspection of the CMDs, itis quite evident that, at any level of magnitude, a good numberof stars redder than the ridge line of NGC 6553 ([Fe/H] =−0.16) is present almost in all fields, and varies from field-to-field depending mostly on galactocentric position.

These objects cannot be accounted for by photometricblends or by any other conceivable photometric effect. Thereis also no indication to suspect that they are not members ofthe M31 system. If this is true, this population of stars doesex-ist in these fields, and should have a very high metal contentgiven its location in the observed CMD.

However, due to the combined effects in the (I,V-I) planeof factors like e.g. (a) the V,I limiting magnitudes, (b) thesat-uration for the brightest stars, (c) the increasing bendingandseparation, and the non-monotonic behaviour of the ridge lineswith increasing metallicity, (d) the possible existence ofpatchydifferential reddening, and (e) the uncertainties in the adoptedmetallicity scale that are larger at the high metallicity end, it isvery difficult to assign a precise value of metallicity to theseobjects. Both the value of their absolute average metallicityand, especially, the detailed distribution over smaller metallic-ity bins are very uncertain and need a much better and deeperanalysis, probably not feasible via photometric means.

Harris et al. (1999) and Harris & Harris (2000, 2002) findthe same result for the halo of NGC 5128. Interestingly, boththe metal poor and metal rich peaks they find in NGC 5128 areidentical to those in the halo of M31.

We emphasize again that any abundance estimate dependson the choice of the calibrating templates. The suggestion ofsub-structures and a fit to the Simple Model of chemical evolu-tion discussed by Harris & Harris (2000, 2002) are tantalizing,but we caution that the present abundances are not sufficientlyaccurate to allow firm astrophysical conclusions with such de-tail.

7. Effect of varying assumptions on the MDs

Metallicity determinations obtained via purely photometricdata (i.e. stellar/population colors) are strongly dependent onvarious assumptions and uncertainties in fundamental issuessuch as (for known/assumed population age) the choice of themetallicity scale, the reddening and the distance modulus.

We explored these effects by re-deriving the MD of a testfield (G64) under a set of different assumptions for the metallic-ity scale, the distance modulus and the reddening. The resultsof this test are shown in Fig. 17. It may be useful, however, todiscuss in some detail the possible impact of each parameter,separately.

7.1. The metallicity scale

Panels (a) and (b) in Fig. 17 show the MD derived for G64 us-ing our standard assumptions, and the CG and ZW metallicityscales, respectively. Our major conclusions are unaffected if wesimply shift by about –0.2 dex (i.e. the offset between the twoscales at mid-metallicity range) the adopted metallicity bound-aries (from -1.4 to -1.6, and from -0.2 to -0.4) when passingfrom CG to ZW.

In particular, we see that both sets of MDs show a poorlypopulated tail reaching as far as [Fe/H] < −2. The use of theZW scale does not affect the median of the MD, which remainsat [Fe/H] ∼ −0.6± 0.1, but it shifts the average by about−0.13dex. A possible additional peak at [Fe/H] ∼ −1.5 dex seemsto emerge. The evidence for the very metal-rich (VMR) com-ponent remains, though shifted by one bin. In fact, a large frac-tion of the VMR objects in the ZW-scale (see figures 18, 19)are found within the interval−0.4 < [Fe/H] < −0.2 due to thenon-linear relationship between the two metallicity scales.


Given the importance of this issue, we have then derived theMDs for all fields using the ZW scale, to ease the comparisonwith the previous studies. Figures 18 and 19 show the resultingdistributions for all the considered fields. We note that:

– the bulk of the population is metal-rich and peaks at[Fe/H] ∼ −0.6, as with CG;

– the young/metal-poor population (MPorY) is well de-tectable and perhaps even enhanced, displaying a slightlybumpy feature whereas the use of the CG scale produces asmoother metal-poor tail;

– a significant very metal-rich population with [Fe/H] >−0.4 is clearly present, though the separation from themetal-rich side of the ”central” component is sligtly lessclear. We also note that a large fraction of these metal-richobjects populate the bin -0.4, -0.2, due the already men-tioned different behaviour of the two metallicity scales inthe very metal-rich regime.

– the percentage of the VMR stars varies from field-to-fieldas with the CG-scale; G219 is confirmed to be peculiar.

In summary, the only possible (and very marginal) differ-ence between the two sets of results is the slightly differentappearance of the MDs in the metal-poor range, that could beinterpreted as an indication for the existence of a possiblesec-ondary peak at [Fe/H]ZW ∼ −1.5 in a few fields. However,we do not attach much weight to this interpretation, since thismight be due to, or enhanced by, a slight discontinuity in thebehavior of the RGB colors as a function of [Fe/H]ZW in thetemplate grid, that occurs at [Fe/H]ZW ∼ –1.4. This discontinu-ity is not present in the CG scale.

We note again that the choice of a different metallicity scaledoes not make any dramatic difference in the results, at least atthe level of detail we believe is compatible with the intrinsicquality of the available data.

7.2. The distance

The panels (c) and (d) in Fig. 17 consider the effect of varyingthe M31 distance modulus by±0.15 mag on the MD of G64.This variation is rather large and must be considered an upperlimit, yet the variation induced on the mean and mode of themetallicity distribution is only±0.07 dex.

Also the variation induced on the relative contributions ofthe three possible components defined by the three metallicityregimes (Sect. 6.2) is quite small.

7.3. The total reddening

The effect of a total reddening variation of±0.05 is not neg-ligible, producing a variation of± 0.13 dex on the mean andmode of the MD, and, consequently, would alter quite signifi-cantly the relative contribution of the smaller components(i.e.MPorY, VMR) to the global abundance distribution. However,the blue locus of the RGB, along with other estimates of thereddening toward M31, constrain the reddening of our fieldswell within this range.

As already noted, since within the adopted approach theMDs are fully drawn from the distributions of the intrinsic col-ors translated into MD via the adopted grid, it is unavoidablethat any systematic shift of the colors correspondingly affectsthe MDs. And, since the relationship between color and metal-licity is highly non-linear, even a small variation of the adoptedreddening (at the level of a few hundredths of a magnitude) in-duces both a shift and a deformation of the MDs, especially inthe metal-poor regime where the sensitivity to color variationis much larger. Note however that the median and mean metal-licity estimates do not vary more than±0.2 dex in response toa±0.05 change in reddening.

7.4. Conclusions from the Tests

The results of the above tests can be summarized as follows:i) With any plausible choice of distance modulus and to-

tal reddening, most of the stars (i.e.∼ 60− 80% of the total)lie in the range−1.4 ≤ [Fe/H] ≤ −0.2 and thepeakof thismain component of the MD lies between [Fe/H] = −0.8 and[Fe/H] = −0.4, independently of the metallicity scale.

ii) Although the existence, as distinct populations, of twoadditional components at the very metal-poor and very metal-rich ends of the MDs may be debatable, the existence of verymetal poor stars (also supported by the presence of a BHB pop-ulation) and of very metal-rich stars seems to be quite firmlyestablished.

iii) As repeatedly noticed, the uncertainties in derivingabundances from photometry are still rather large, and settlingthe question of whether the abundance distributions are merelyskewed or bimodal or even tri-modal will require much moreaccurate and reliable means of metallicity determinations, e.g.better photometric data coupled with a more reliable and ex-tended grid of reference GC or the availability of spectroscopicdata for a wide sample of stars.

8. Comparison with other MDs in the literature

Several authors have investigated M31 fields using bothground-based facilities andHS T observations. We here com-pare our present results with those of the most recent previousanalyses, noting that all of them work in the ZW metallicityscale.

– Holland et al. (1996)These authors studied the halo fields near G302 and G312,located 32 and 50 arcmin approximately along the SE mi-nor axis, respectively. The reddening and distance modu-lus assumed for M31 wereE(B − V) = 0.08± 0.02 and(m − M)0 = 24.3 ± 0.1. The RGBs were compared withthe RGB ridge lines for three Galactic GCs (i.e. M15,NGC1851 and 47 Tuc) and two metal-rich fiducials se-lected from isochrones with aget0 = 13.8 Gyr and [m/H]=–0.4 and 0.0. The resulting MDs show a spread in metallicityof −2 ≤ [m/H] ≤ −0.2 with the majority of stars having[m/H] ∼ −0.6.We have not analysed these fields, but we can comparethese results with ours on halo fields such as G105 and


G319. The (V − I )0 color distribution of the RGBs at theapproximate level I∼ 22.45 (i.e.MI = −2.0) seems to peakaround 1.25± 0.2, at a cursory inspection of their Fig. 2and 3; the resulting MDs are in very good agreement withour results, both in the shape and in the location of the MDpeak at [Fe/H] ∼ −0.6.

– Sarajedini & Van Duyne (2001)The disk-dominated field near G272 was recently analysedby Sarajedini & Van Duyne (2001), who derived a MD forit. Their analysis is based on the assumptions thatE(B −V) = 0.08 and (m− M)0 = 24.5 for M31. Their fiducialsequences are the RGB ridge lines of 6 Galactic globularclusters and one open cluster, of which only 47 Tuc is incommon with our grid.The interpolation procedure, similar in principle to ours,was applied within a narrow range of magnitude along theRGB, i.e. I=22.65± 0.1 corresponding toMI = −2.0± 0.1mag. The (V − I )0 color distribution within this strip (Fig.6 in Sarajedini & Van Duyne (2001)) has a Gaussian shapepeaking at (V− I )0 ∼ 1.4, whereas the color distribution wederive from our data under our assumptions peaks at (V −I )0 ∼ 1.2. This difference is largely accounted for by thedifferent choice of reddening; differences in the absolutecalibration (e.g., aperture corrections) may also contribute.The MD obtained by Sarajedini & Van Duyne (2001) issimilar in shape to ours, with an extended metal-poor tailreaching [Fe/H] ≤ –2, but their main peak at [Fe/H] ∼ –0.2 is definitely shifted by about 0.2 dex toward the metal-rich end. This is at least partly due to the quoted systematicdifference in the color distribution.On this basis, these authors conclude that their Gaussiancomponent peaking at< [Fe/H] >= −0.22± 0.26 com-prises 70 % of the total number of stars in the sample, andis attributed to the thick-disk population.We cannot compare this result homogeneously with oursbecause we have not fitted our MD with Gaussian compo-nents: however, we note that the fraction of very metal-richstars with [Fe/H] ≥ –0.4 is∼ 10% in the ZW scale (see Fig.18 and Table 5), which is not unusually large and is quiteconsistent with the position of G272 with respect to the ma-jor axis of M31. If indeed we are observing the thick-diskpopulation in this field, its relative contribution is not over-whelming and is found also, and to a larger extent, in all theother fields located on the disk of M31 (see Sect. 9).

– Ferguson & Johnson (2001)The far outer disk field near G327 was recently investi-gated by Ferguson & Johnson (2001). Under the standardassumptions of E(V–I)=0.10 and (m−M)0=24.47, the RGBwas compared with the fiducial ridge lines of three GGCs inthe metallicity range –1.9 to –0.3. A best fit was performedat the luminosity levelMI = −0.13± 0.1 mag, consistentwith a predominantly old-to-intermediate age stellar pop-ulation with [Fe/H] ∼ –0.7 plus a trace population of oldmetal-poor stars.In this field we also find a MD with a median value at[Fe/H] ∼ –0.66 (ZW-scale), and only∼ 20% of stars with[Fe/H] < –0.8.

– Durrell et al. (2001)

An outer halo field located 90 arcmin (i.e. 20 kpc) SE of theM31 nucleus and roughly along the minor axis was studiedby Durrell et al. (2001) using CFHT V,I data. A reddeningvalueE(V − I ) = 0.10± 0.02 was derived from the colordistribution of the foreground Milky Way halo stars, and adistance modulus (m − M)0 = 24.47± 0.12 was derivedfrom the luminosity of the RGB tip.A cursory inspection of their Fig. 7 indicates that the RGBstars at V∼ 22.6 (corresponding toMI ∼ −2) populate a(V–I) color range between approximately 0.9 and 1.6 witha presumable accumulation between 1.2 and 1.4.The MD functions are derived by comparison with evolu-tionary tracks for 0.8M⊙ stars. The MD for the RGB mag-nitude interval 20.6< I < 22.5 is skewed toward the metal-poor side reaching almost [m/H]=–2.5, and could be wellfitted by two Gaussian components peaking at [m/H]=–0.52 and –1.20 and including 60% and 40% of the total stel-lar population, respectively. The agreement with our anal-ysis of the distant fields G105 and G319 is quite satisfying(see Tables 4 & 5).

– Reitzel & Guhathakurta (2002)A spectroscopic study based on Keck-LRIS data of about30 halo red giant stars in a field at R=19 kpc on the SEminor axis of M31 was done by Reitzel & Guhathakurta(2002). The MD they find for these halo giants spans morethan 2 dex range with a mean/median value [Fe/H] ∼ –1.9to –1.1 (depending on calibration and sample selection).However, the high metallicity end of this distribution ispoorly constrained by these data since the selection func-tion for secure M31 members excludes> 80% of the giantsin solar/super-solar metallicity range.No direct comparison is possible with our results, exceptfor a general comment on the confirmation of the existenceof a well-detectable fraction of metal-poor stars in the outerhalo of M31 (see Sect. 6.2). Our field G319 is within∼ 30arcmin of their field on the minor axis, and we do observethe MR population found in other fields.

9. Correlation of metallicity with position

9.1. The wide-field survey by Ferguson et al. (2002)

In a very important photometric survey of the halo and outerdisk (appeared when the present analysis was nearly com-pleted), Ferguson et al. (2002) have studied in great detailboththe spatial density and metallicity variations (as inferred fromcolor information, like in the present paper), covering an areaof about 25 square degrees around M31. This is by far the mostdetailed and complete ground-based study carried out so faronthe field population in M31.

Since the CCD data were obtained with the 2.5m INT-WFC at La Palma, the limiting magnitudes (withS/N = 5)are V=24.5 andi = 23.5, much brighter than those we haveobtained from ourHS T photometry. The shallower limits inmagnitude are however amply balanced by the much wider areasampled. Therefore this survey and our work fromHS T dataare nicely complementary.


Their conclusions, very convincingly supported by their fig.2-7, are schematically:

– There is evidence for both spatial density and metallicityvariations across the whole body of M31, which are often,but not always, correlated.

– Besides the known Ibata’s stream, two other overdensitiesat large radii can be detected, close to the SW major axis, inthe proximity of the very luminous GC G1 (called the ”G1-clump”), and near the NE major axis, coinciding with andextending beyond the previously known ”northern spur”.

– The most prominent metallicity variations are found in twolarge structures in the southern half of the halo, the firstone coinciding with the giant stellar stream found by Ibataet al. (2001), and the second one corresponding to a muchlower stellar overdensity. Their metallicities are above theaverage value, which corresponds approximately to that of47 Tuc, i.e. [Fe/H] ∼ −0.7.

Since the areas covered by the previous studies are verysmall compared to this wide survey, as correctly pointed outby Ferguson et al. there is no ground to identify significantconflicts between the results of previous studies on individ-ual fields and those derived from this general panoramic study.This is especially true because of the clear detection of streamsand substructures (both in star density and metallicity) alloverM31, which has made somehow ”unpredictable” the nature ofthe population one would expect to find on the basis of the mereX,Y location within M31.

This fact has an obvious impact on the general interpreta-tion we might derive from our analysis, as the lines of sight wehave investigated, though quite numerous (16+ G1), could benot sufficiently representative of the global scenario.

In this new light we have reconsidered the properties ofour observed fields, with particular attention to themap of thespatial and chemical substructuresFerguson et al. (2002) pre-sented in their Fig. 7, and reported here in Fig. 20 where thelocations of our fields are overplotted. We anticipate that theresults coming out from our deeperHS Tfields suggest that in-deed some of the considered lines of sight may be affected bythe presence of streams and overdensities.

9.2. Our 16+1 HS T fields

As reported in figures 14 and 15 and Table 4, a number of pa-rameters have been derived from the metallicity distributionsfor each field. A straightforward application is to check firstfor any trend with position in the galaxy, recalling the alreadyquoted caveat that only projected distances are available.

9.2.1. The (X,Y)–plane

Fig. 21a, b, c report the fractions of MPorY and VMR stars asa function of the X and Y coordinates and galactocentric dis-tance R in arcmin, respectively. The error bars associated toeach point are based on the square roots of the star numbers ineach individual sample.

Because of the large difference in the error bars associatedto the different fields, as the respective population size ranges

from about 10,000 stars in the inner fields down to less than100 in the outer halo fields, it is quite hard to assign a fair sta-tistical significance to any trend one might see. Nevertheless,considering also the histograms presented in figures 14 and 15,some general comments can be made:

– The fraction of MPorY stars is substantially constant andindependent of position. As already noted, this does notnecessarily mean that there is a constant percentage ofmetal-poor halo stars in all the observed fields, because ofthe possible (and, at the present status, unknown) contam-ination by Young objects in the inner samples. However,metal-poor old stars are surely present in the halo, albeit inrather small quantity.

– The inner fields (G87, G287) have a population with ahigher average metallicity than all other fields, resultingfrom the combination of a higher than average VMR frac-tion and a (correspondingly) smaller MR fraction (remem-ber that the total MPorY+ MR + VMR = 1.0). This evi-dence is especially clear looking at the Y-plot (Fig. 21b).

– Ferguson et al. (2002) noted that the average metallicityof the SW fields (i.e. with negative Y and negative X) isslightly larger than the average metallicity estimated fortheNE fields. Though the statistical significance is weak, thiseffect is perhaps visible also in Fig. 21a,b, as a slight en-hancement of the mean fraction of VMR stars in theY < 0andX < 0 regions.

– G219 and G319 have an unusually large fraction of metal-rich stars for their distance from the M31 center (90 and 70arcmin respectively). This anomaly may be connected withtheir (projected) proximity to the Ferguson et al. (2002)stream.

In summary, the considerations reported above show that asimple description of the results making use of just theHS Tfields and their projected location in the X.Y-plane coupledwith a ”simple” description of the M31 outer disk and halomay be not fully capable to extract all the information poten-tially offered by the available data.

This has prompted us to look at our result within the newscenario emerging from Ferguson et al. (2002) work.

9.2.2. HST fields and detected substructures

If we briefly rediscuss our data within the framework reportedin Fig. 20 (where the fields are identified over the cartoon re-produced from Fig. 7 of Ferguson et al.), we can add somefurther notes:

– Several fields (e.g. G33, G219, G105, G58, G64, G108) arelocated or projected on the ”giant stream” first identified byIbata et al. (2001) and now confirmed by Ferguson et al.(2002).Some of these fields (see Table 4,5) show a very high frac-tion of VMR stars, with a clear peak in the MD. The G219field presents one of the best examples of this case. G219 islocated well off the plane but lies precisely on the detectedstream; the field has an anomalously large fraction (∼ 20%)of very metal rich stars.


Though one cannot conclude for sure that the existence ofthese VMR stars in most of these fields has to be ascribed tothe possible physical connection with the ”giant stream”, itseems conceivable that the two facts may be related. On theother hand, there are also fields that are located/projectedon the stream which do not show a similarly evident ”ex-cess” of VMR stars (see f.i. G58, G64, G105).

– The MD of G1 may also be worth of a short commentwithin the new scenario. In fact, even though it is a distantfield (projected some 32 kpc from the nucleus) and with atotal population of measured stars of only 44 objects, itsMD shows (see Fig. 16) a weak indication for a secondmetal-rich peak (at [Fe/H] ∼ −0.2, with F(VMR)∼ 10%.Since, as shown in the map in Fig. 20, the region called byFerguson et al. the ”G1 clump” denotes a possible substruc-ture (at least in star density) within the area they surveyed,it may be of some relevance to investigate further the exis-tence and origin of VMR objects in this field.

In summary, the re-analysis of the properties of the MDswithin the framework provided by Ferguson et al. (2002) doesnot yield afully satisfactory ordering and detailed explanationfor all the fields investigated in the present study. However,some (otherwise unusual) results are nicely fitting (or, at least,are compatible with) a scenario that accounts for the existenceof significant substructures within M31. These may add an im-portant piece of information for a detailed description of theformation and evolution of this galaxy, including the possibleinteraction with close companions such as M32 (Bekki et al.,2002; Choi et al., 2002).

9.3. Large-scale trends

In the previous sections we have carried out and discussed theanalysis of the MDs we have obtained from the histogramsshown in figures 14 and 15 assuming that, independently ofthe physical reasons or adopted models for M31, one couldidentify two features or discontinuities, at [Fe/H] = −1.4 and[Fe/H] = −0.2, to separate the stars into three main groups. Weare quite confident that such an approach is meaningful and theconsiderations listed above add some support to this.

An alternative and more schematic approach could be sim-ply to consider two groups of stars obtained by dividing thetotal MDs for instance at [Fe/H] = −0.8.

This value for the cut has been chosen to identify roughlywhat we call hereafter the “Metal-Rich Population” (MRP)from the “Metal-Poor Population” (MPP) on the basis of thealready quoted (weak) evidence of a small discontinuity inthe MDs at that bin-border (deduced from inspecting the his-tograms, see f.i. G33, G322, G108, G58, G351, G219, G327;see also Fig. 12).

Table 5 reports for each field the values so obtained fromthe histograms for the fraction of the Metal-Poor Population, F(MPP), and the Metal-Rich Population, F(MRP), the Y-coordinate, and the average (median) metallicity of the totalsample.

Fig. 22 a,b shows the plots of F(MPP) and of the medianmetallicity vs. the absolute Y-coordinate, respectively.

As said, the contribution of stars from the spheroid is ex-pected to grow faster along the Y-direction because of the in-clination of the M31 disk with respect to the plane of the sky.

In fact, F(MPP) grows very rapidly from 0.10 at|Yarcmin| ∼0, to∼ 0.45 at |Yarcmin| ∼ 17, showing a good correlation be-tween the involved variables. However, from|Yarcmin| ∼ 20 to|Yarcmin| ∼ 70, F(MPP) remains nearly constant at F(MPP)∼0.4, with a larger scatter. Note that the two “outliers” with lowvalues of F(MPP) at quite large Y are G219 and G351, bothlocated on or near the ”giant stream” (see Fig. 20). A similarbehaviour is shown by the median metallicity (Fig. 22b)

The trend displayed in figures 22a and 22b is quite clear:the MRP dominates the inner region but the relative importanceof the MPP population rapidly grows going far from the visibledisk. However, in the explored range of|Yarcmin|, the Metal-Poorpopulation never becomes dominant, Metal-Rich stars remainthe major component of the stellar mix.

In summary, these plots suggest that, besides the existenceof important substructures, there is also evidence of a morereg-ular pattern in the inner parts of the galaxy, possibly testifyingof a relatively homogeneous and ordered formation process,later perturbed by merging events.

10. Summary and Conclusions

We have analysed HST-WFPC2 images of 16 fields in M31 ata wide range of distances from the galactic center and plane.These fields have been observed in the F555W (V) and F814W(I) filters and reach a roughly uniform depth.

The color-magnitude diagrams of these fields are generallydominated by a red giant branch and a populous red clump.The RGB has a strong descending tip, indicative of a metal richpopulation. In the outermost fields a blue Horizontal Branchisalso detected. When present, the blue HB represents about 15%of the total HB population, roughly estimated from the fractionof metal poor giants.

In fields superposed on the disk, a blue plume of main se-quence stars is also identified. We report a simple analysis ofthe blue plume population, which shows that the star formationhistory of the disk has been spatially inhomogenous during thelast 0.5 - 1 Gyr.

We have obtained photometric metallicity distributionsfrom RGB stars by interpolation on a grid of empirical glob-ular cluster RGB templates.

The most robust result of the presentHS T survey is theevidence that the metal-rich population, with [Fe/H] ∼ −0.6,is the major component of the stellar mix everywhere in the10 ≤ Rarcmin ≤ 130 range. A minor metal-poor component(with [Fe/H] < −1.4) is also ubiquitous. The old stellar popu-lation is remarkably uniform across the disk, outer halo fields,and in the proposed tidal stream. This uniformity is the domi-nant feature of the old stellar population.

This basic result is not new (see, e.g., Durrell et al., 2001,and references therein), but coupled with data shown by thevery wide mapping carried out by Ferguson et al. (2002), fixeson solid grounds the conclusion that the stellar populationofthe M31 spheroid differs substantially from that of the MilkyWay and is almost an order of magnitude more metal rich, on


average. In this respect, we confirm also the finding of Richet al. (1996), that the metal rich population is present evenat150 arcmin (∼ 35 kpc) from the nucleus, at the location of theglobular cluster G1.

An obvious explanation for the origin of the metal rich haloof M31 is that the population formed in connection with theoutflow of metal-enriched winds, perhaps associated with theformation of the spheroid. Metal rich populations in halos areclearly widespread (e.g. NGC 5128; Soria et al., 1997; Harriset al., 1999; Harris & Harris, 2002). There is evidence for theoutflow of metal enriched winds in Lyman break galaxies, andfor near-solar abundances in these galaxies, at high redshift(Steidel et al., 1996). The stars formed from this material mightwell live to comprise the population II halo. The outflow of en-riched material will not affect the shape of the abundance distri-bution (the classic Simple Model form is retained) but the yield(mean abundance of the whole galaxy) does decline (Hartwick, 1976).

The increasing fraction of the ”Metal-Poor Population” (i.e.with [Fe/H] < −0.8) with distance from the plane Y suggests adissipational origin for the regions that are closer to the galacticplane. This evidence is compatible with the hypothesis thata”homogeneous” formation event may have occurred in the veryearly stages of the M31 history.

On the other hand, the image of the M31 halo illustratedby Ibata et al. (2001) and Ferguson et al. (2002) shows (for themetal-rich stars taken as aglobalpopulation) a quite flattened,disk-shaped distribution, associated with the outskirts of the in-ner disk. It is possible thus that the metal rich population mightbe associated with a proto disk or flattened halo. The largenumber of stars involved would strongly favor the notion thatwe study theold disk, and not just some trace population suchas the thick disk. So we have this alternative scenario, wherethe dominant metal-rich population might be associated with ameta-disk extending perhaps more the 20 kpc, while the metal-poor population might be associated with the spheroid. The twopopulations – and the corresponding structures – would havedifferent evolutionary histories. The old stellar disk would bemuch larger than previously believed, extending out to a fewdegrees from the nucleus of M31 and dominating the stellarmix also along lines of sight very distant from the center of thegalaxy (Ferguson & Johnson, 2001; Ferguson et al., 2002). Itis also possible that the inner disk and bulge had common, orclosely tied, formation histories.

In addition to this, the sub-structures (“streams” and“clumps”) clearly detected by Ibata et al. (2001); Fergusonetal. (2002) and supported by our results might be related to thephenomena of merging or interaction with satellites. In thiscase the metal rich component might be connected with thetidal disintegration of a companion (Ferguson et al. (2002)), alikely candidate being M32, perhaps involving a good deal ofits mass (Bekki et al., 2002; Choi et al., 2002).

New data are now required to choose among the scenar-ios presented here. Deeper HST imaging, some of which is inprogress, may constrain the actual age distribution of stars insome regions of the halo. Large scale spectroscopic surveysofhalo stars have just begun, and have the potential to determinewhat fraction of the M31 halo might have originated from the

tidal disruption of satellites. Our first round of large-scale HSTimaging gives evidence that the old “halo” stellar population ismore metal rich than that of the Milky Way, with surprisinglylittle variation in the properties of the old stellar population,even for fields ranging up to galactocentric distances exceed-ing 30 kpc. The challenge now will be assemble a data set forthe Galaxy and M31 powerful enough to constrain their origins.

Acknowledgements.We are indebted to Helmut Meusinger for kindlyproviding us with the plates of the Tautenburg Schmidt Telescope,and to Roberto Merighi for help in drawing some figures. Useful dis-cussions with Gisella Clementini, Carlo Corsi, George Djorgovski,Francesco Ferraro, Wendy Freedman, Puragra Guhathakurta,MonicaTosi are also kindly acknowledged. Support for Michael Rich’s activ-ities on proposal GO-6671 was provided by NASA trhough a grantfrom the Space Telescope Science Institute, which is operated bythe Association of Universities for Research in Astronomy,Inc., un-der NASA contract NAS 5-2655. Grants: ASI J/R/35/00, MURSTMM02241491-004.

References

Aloisi, A., Tosi, M., & Greggio, L. 1999, AJ, 118, 302Baade, W. 1944, ApJ, 100, 137Barmby, P., Huchra, J.P., Brodie, J.P., Forbes, D.A., Schroder, L.L., &

Grillmair, C.J. 2000, AJ, 119, 727Battistini, P., Bonoli, F., Braccesi, A., Federici, L., Fusi Pecci, F.,

Marano, B., & Borngen, F. 1987, A&AS, 67, 447Bekki, K., Couch, W.J., Drinkwater, M.J., & Gregg, M.D. 2002ApJ

557, L39Bertelli, G., Bressan, A., Chiosi, C., Fagotto, F., & Nasi, E. 1994,

A&AS, 106, 275Bertin, E., & Arnouts, S., A&AS, 117, 393Biretta J. (ed.), 1996, The WFPC2 Instrument Handbook, Baltimore,

STScIBurstein, D., & Heiles, C. 1982, AJ, 87, 1165Carretta, E. & Gratton, R. 1997, A&AS, 121,95Carretta, E., Cohen, J.G., Gratton, R.G., & Behr B.B. 2002, AJ, 122,

1469Choi, P.I., Guhathakurta, P., & Johston, K.V. 2002, AJ, 124,310Cohen, J.G., Gratton, R.J., Behr, B.B., & Carretta, E. 1999,ApJ,523,

739Couture, J., Racine, R., Harris, W.E., & Holland, S. 1995, AJ, 109,

2050Da Costa, G.S., Armandroff, T.E., & Caldwell, N. 2002, AJ, 124, 332Dean, J.F., Warren, P.R., & Cousins, A.W.J., 1978, MNRAS,183, 569Vaucouleurs, G. de, 1958, ApJ, 128, 465Dolphin, A.E. 2000, PASP, 112, 1397Durrell, P.R., Harris, W.E., & Pritchet, C.J 1994, AJ, 108, 2114Durrell, P.R., Harris, W.E., & Pritchet, C.J. 2001, AJ, 121,2557Ferguson, A.M.N. & Johnson, R.A. 2001, ApJ, 559, L13Ferguson, A.M.N., Irwin, M.J., Ibata, R.A., Lewis, G.F. & Tanvir, N.R.

2002, AJ, 124, 1452Ferraro, F.R., Messineo, M., Fusi Pecci, F., de Palo, M.A., Straniero,

O., Chieffi, A., & Limongi, M. 1999, AJ, 118, 1738Freedman, W.L., & Madore, B.F. 1990, ApJ, 365, 186Gallart, C., Freedman, W.L., Aparicio, A., Bertelli, G., & Chiosi, C.

1999, AJ, 118, 2245Guarnieri, M.D., Renzini, A., & Ortolani, S. 1997, ApJ, 477,L21Guhathakurta, P., Reitzel, D.B., & Grebel, E. 2000, in SPIE

Conference: Discoveries and Research Prospects with 8-10 MeterClass Telescopes, in press (astro-ph/004371)


Harris, G.L.H., Harris, W.E. & Poole, G.B. 1999, AJ, 117, 885Harris, G.L.H. & Harris, W.E. 2000, AJ, 120, 2423Harris, W.E. & Harris, G.L.H. 2002, AJ, 123, 3108Hartwick, F.D.A. 1976, ApJ, 209, 418Hernandez, X., Valls-Gabaud, D., & Gilmore, G. 1999, MNRAS,304,

705Hodge, P.W. 1979, AJ, 84, 744Hodge, P.W. 1992, The Andromeda Galaxy, Astrophysics and Space

Science Library, vol 176, Dordrecht, KluwerHolland, S. 1998, AJ, 115, 1916Holland, S., Fahlman, G.G., & Richer, H.B. 1996, AJ, 112, 1035Holland, S., Fahlman, G.G., & Richer, H.B. 1997, AJ, 114, 1488Holtzman, J.A., Burrows, C.J., Casertano, S., Hester, J.J., Trauger,

J.T., Watson, A.M., & Worthey, G. 1995, PASP, 107, 1065Ibata, R., Irwin, M., Lewis, G., Ferguson, A.M.N., & Tanvir,N. 2001,

Nature, 412, 49Jablonka, P., Bridges, T.J., Sarajedini, A., Meylan, G., Maeder, A., &

Meynet, G. 1999, ApJ, 518, 627Jablonka, P., Courbin, F., Meylan, G., Sarajedini, A., Bridges, T.J., &

Magain, P. 2000, A&A, 359, 131McClure, R., & Racine, R. 1969, AJ, 74, 1000Morris, P.W., Reid, I.N., Griffiths, W.K., & Penny, A.J. 1994,

MNRAS, 271, 852Mould, J., & Kristian, J. 1986, ApJ, 305, 591Olsen, K.A.G., Hodge, P.W., Mateo, M., Olszewski, E.W., Schommer,

R.A., Suntzeff, N.B., & Walker, A.R. 1998, MNRAS, 300, 665Ortolani, S., Renzini, A., Gilmozzi, R., Marconi, G., Barbuy, B., Bica,

E., & Rich, R.M. 1995, Nature, 377, 701Perrett, K.M., Bridges, T.J., Hanes, D.A., Irwin, M.J., Brodie, J.P.,

Carter, D., Huchra, J.P., & Watson, F.G. 2002, AJ, 123, 2490Pritchet, C.J., & van den Bergh, S. 1994, AJ, 107, 1730Reitzel, D.B., & Guhathakurta, P. 2002, AJ, 124, 234Renzini, A. 1998, AJ, 115, 2459Renzini, A., & Buzzoni, A. 1986, in Spectral evolution of galaxies,

Dordrecht, Reidel, p. 195Rich, R.M., Mighell, K.J., Freedman, W.L., & Neill, J.D. 1996, AJ,

111, 768Rich, R.M., Corsi, C.E., Federici, L., Fusi Pecci, F. , & Cacciari, C.

2002, in preparationRicher, H.D., Crabtree, D.R., & Pritchet, C.J 1990, ApJ, 335, 448Sarajedini, A. & Van Duyne, J. 2001, AJ, 122, 2444Sargent, W.L.W., Kowal, C.T., Hartwick, F.D.A., & van den Bergh, S.

1977, AJ, 82, 947Saviane, I., Rosenberg, A., Piotto, G., & Aparicio, A. 2000,A&A,

335, 966Schechter, P., Mateo, M., & Saha, A. 1993, PASP, 105, 1342Simien, F., Athanassoula, E., Pellet, A., Monnet, G., Maucherat, A.,

& Courtes, G. 1978, A&A, 67, 73Soria, R. et al. 1996, ApJ, 465, 79Steidel, C.C., Giavalisco, M., Pettini, M., Dickinson, M.,&

Adelberger, K.L. 1996, ApJ, 462, L17Stephens, A.W., et al., 2001, AJ, 121, 2597Tinsley, B.M., & Spinrad, H. 1971, Ap&SS, 12, 118Tosi, M., Greggio, L., & Focardi, P. 1989, Ap&SS, 156, 295van den Bergh, S. 2000, The galaxies of the Local Group, Cambridge

University Press, p. 11Walker, A.R. 1994, AJ, 108, 555Walterbos, R.A.M., & Kennicutt, R.C.Jr. 1988, A&A, 198, 61Whitmore, B., Heyer, I., & Casertano, S. 1999, PASP, 111, 1559Williams, B.F. 2002, MNRAS, 331, 293Zinn, R.J., & West, M.J. 1984, ApJS, 55, 45


Table 1.Location of the observed fields.

G Bo Xarcmin Yarcmin Rarcmin E(B-V) Fsph

G1 −149.54 29.88 152.50 0.11 > 0.9G11 293 −61.72 43.91 75.75 0.11 > 0.9G33 311 −57.58 1.24 57.59 0.11 0.07G58 6 −6.80 27.37 28.20 0.11 0.70G64 12 −10.65 23.01 25.35 0.11 0.65G76 338 −44.15 −8.84 45.03 0.14 < 0.1G87 27 −26.41 1.00 26.43 0.14 0.18G105 343 −57.57 −29.81 64.83 0.11 0.80G108 45 7.38 20.21 21.52 0.11 0.60G119 58 −28.88 −10.06 30.59 0.14 < 0.1G219 358 −64.79 −58.32 87.17 0.11 > 0.9G272 218 17.13 −16.88 24.05 0.14 < 0.1G287 233 35.44 −0.26 35.45 0.14 0.12G319 384 −21.18 −68.89 72.07 0.11 0.99G322 386 61.64 −4.49 61.80 0.11 0.07G327 130.01 5.10 130.11 0.11 > 0.9G351 405 63.14 −53.71 82.90 0.11 > 0.9


Table 2.Observations.

Name R.A. (2000) Dec. (2000) Date Proposal P.I. Filters (texp[s])

G11 0 36 20.78 40 53 36.60 Feb 3,2000 GO 6671 Rich F814W (5400),F555W (5300)G33 0 39 33.75 40 31 14.36 Feb 26,1999 GO 6671 Rich F814W (5400), F555W (5300)G64 0 40 32.79 41 21 44.65 Aug 17,1999 GO 6671 Rich F814W (5400), F555W (5300)G76 0 40 58.83 40 35 47.32 Jan 11,1999 GO 6671 Rich F814W (5400), F555W (5300)G87 0 41 14.61 40 55 51.12 Aug 16,1999 GO 6671 Rich F814W (5400), F555W (5300)G119 0 41 53.01 40 47 08.63 Jun 13,1999 GO 6671 Rich F814W (5400), F555W (5300)G287a 0 44 41.97 41 43 56.40 Sep 26,1999 GO 6671 Rich F814W (5400), F555W (5300)G287b 0 44 41.97 41 43 56.40 Sep 26,1999 GO 6671 Rich F814W (5400), F555W (5300)G319 0 46 21.92 40 17 00.01 Feb 28,1999 GO 6671 Rich F814W (5400), F555W (5300)G322 0 46 26.94 42 01 52.94 Jan 10,1999 GO 6671 Rich F814W (5400), F555W (5300)G327 0 49 38.91 43 01 13.85 Jun 19,1999 GO 6671 Rich F814W (5400), F555W (5300)

G272 0 44 51.40 41 19 16.00 Jan 22,1995 GO 5420 Fusi Pecci F814W(10800), F555W (3800)G351 0 49 58.10 41 32 17.00 Jan 18/19,1995 GO 5420 Fusi Pecci F814W(10800), F555W (3800)

G58 0 40 26.79 41 27 27.72 Feb 15,1994 GTO 5112 Westphal F814W (2000), F555W (2000)G108 0 41 43.26 41 34 20.76 Feb 15,1994 GTO 5112 Westphal F814W(2000), F555W (2000)G105 0 41 43.17 40 12 22.76 Feb 15,1994 GTO 5112 Westphal F814W(2000), F555W (2000)G219 0 43 17.81 39 49 13.53 Feb 15,1994 GTO 5112 Westphal F814W(2000), F555W (2000)


Table 3.Grid of Galactic Globular Clusters template for metal-licity determinations in the Zinn & West (1984) and Carretta&Gratton (1997) scales

[Fe/H]ZW [Fe/H]CG E(B-V) (m− M)0

NGC6341 M92 −2.24 −2.16 0.02 14.74NGC6205 M13 −1.65 −1.39 0.02 14.38NGC5904 M5 −1.40 −1.11 0.03 14.31NGC104 47Tuc −0.70 −0.70 0.04 13.29NGC6553 −0.34 −0.16 0.84 13.44NGC6528 −0.23 0.07 0.62 14.35


Table 4. Metallicity Distributions in the Carretta & Gratton (1997)scale: population (in percentage) of individual metallicitybins (step 0.2 dex) for each observed field, and for all (coadded) fields.

Name < −2.5 −2.5 −2.3 −2.1 −1.9 −1.7 −1.5 −1.3 −1.1 −0.9 −0.7 −0.5 −0.3 −0.1 0.1 .2− .27 > .27

G1 0.0 2.3 0.0 0.0 0.0 0.0 2.3 11.4 11.4 22.7 20.5 4.5 9.1 9.1 2.3 2.3 2.7G11 0.0 1.1 0.0 0.0 0.6 0.6 1.1 2.2 9.0 17.4 24.7 23.0 9.0 9.0 1.1 0.0 1.2G33 1.3 0.2 0.2 0.4 0.6 0.7 0.4 3.5 5.8 8.2 15.2 22.5 16.1 19.0 4.8 0.5 0.6G58 0.4 0.1 0.5 0.5 0.9 1.1 1.2 5.9 11.2 15.5 25.6 23.7 9.5 3.3 0.5 0.0 0.0G64 0.6 0.2 0.2 0.4 0.4 0.8 0.5 6.0 8.6 14.7 21.5 25.6 11.5 7.3 1.3 0.1 0.5G76 3.6 0.3 0.6 0.7 0.9 0.9 0.9 3.7 5.3 8.5 13.6 21.8 15.7 16.7 4.4 0.6 1.7G87 0.4 0.0 0.2 0.2 0.3 0.3 0.4 1.7 3.2 5.2 10.3 16.5 15.1 26.2 13.2 2.3 4.6G105 0.2 0.0 0.0 0.7 0.4 0.2 1.1 5.0 8.8 15.3 21.9 25.8 12.3 6.8 0.9 0.2 0.4G108 1.2 0.2 0.4 0.4 0.5 0.9 0.8 5.0 9.0 14.0 22.2 24.4 12.0 7.7 1.0 0.1 0.3G119 3.9 0.3 0.8 0.9 1.2 1.0 1.3 5.2 7.8 10.4 14.8 22.1 14.6 11.62.8 0.5 0.8G219 0.8 0.0 0.0 0.0 0.8 1.6 2.4 2.4 2.4 7.1 17.3 26.0 18.9 18.1 2.4 0.0 0.0G272 4.1 0.3 0.6 1.1 1.5 2.0 1.6 7.9 10.5 13.2 17.3 20.7 11.0 6.41.3 0.2 0.4G287 0.5 0.1 0.2 0.3 0.3 0.4 0.4 2.3 4.4 7.9 13.9 24.3 18.6 20.7 4.3 0.6 0.8G319 0.0 0.0 2.4 0.0 0.0 1.2 2.4 4.8 11.9 13.1 17.9 11.9 13.1 11.9 6.0 0.0 3.6G322 1.8 0.2 0.3 0.5 0.7 0.3 0.6 3.4 6.2 10.0 17.2 21.9 16.5 15.73.7 0.3 0.5G327 0.0 0.0 2.4 0.0 0.0 0.0 0.0 1.2 8.3 10.7 33.3 19.0 13.1 9.5 1.2 0.0 1.2G351 0.8 0.0 0.8 0.0 0.8 0.0 0.0 5.9 7.6 9.3 21.2 26.3 16.1 5.1 2.5 0.8 2.5ALL 1.7 0.2 0.4 0.5 0.7 0.7 0.7 3.7 6.0 9.2 15.0 21.4 15.0 16.8 5.5 0.9 1.7


Table 5.Parameters derived from the Metallicity Distributions in the Carretta & Gratton (1997) scale. Columns: (1) name of thefield, (2) fraction of stars with [Fe/H] < −1.4 [Metal-Poor or Young – MP or Y], (3) with−1.4 < [Fe/H] < −0.2 [Metal-Rich–MR], (4) with [Fe/H] > −0.2 [Very Metal-Rich – VMR] (5) fraction of stars with [Fe/H] < −0.8 [Population Metal-Poor –PMP], (6) fraction of stars with [Fe/H] > −0.8 [Population Metal-Rich –PMR], (7) average metallicity ([Fe/H]ave) (8) associatedstandard deviation, (9) median metallicity, (10)associated semi-interquartile interval. The corresponding figuresin the Zinn &West (1984) scale are partially reported in Fig. 18 and 19.

Name Fr(MP or Y) Fr(MR) Fr(VMR) Fr(PMP) Fr(PMR) [Fe/H]av σav [Fe/H]med σmed

G1 0.045 0.795 0.136 0.500 0.477 -0.79 0.49 -0.82 0.26G11 0.034 0.854 0.101 0.320 0.669 -0.68 0.38 -0.64 0.20G33 0.026 0.713 0.242 0.200 0.781 -0.52 0.42 -0.48 0.25G58 0.043 0.915 0.038 0.369 0.627 -0.75 0.37 -0.67 0.22G64 0.025 0.878 0.087 0.317 0.672 -0.68 0.37 -0.63 0.22G76 0.043 0.686 0.218 0.218 0.729 -0.55 0.46 -0.49 0.26G87 0.013 0.520 0.418 0.114 0.837 -0.35 0.39 -0.28 0.25G105 0.024 0.891 0.079 0.315 0.678 -0.67 0.35 -0.63 0.22G108 0.031 0.866 0.087 0.311 0.673 -0.68 0.38 -0.63 0.22G119 0.056 0.749 0.149 0.289 0.664 -0.65 0.47 -0.57 0.28G219 0.047 0.740 0.205 0.165 0.827 -0.53 0.38 -0.50 0.22G287 0.017 0.713 0.256 0.163 0.568 -0.48 0.37 -0.44 0.24G272 0.071 0.805 0.079 0.386 0.824 -0.76 0.46 -0.67 0.28G319 0.060 0.726 0.179 0.357 0.607 -0.67 0.49 -0.63 0.30G322 0.026 0.713 0.242 0.224 0.753 -0.56 0.41 -0.52 0.25G327 0.024 0.857 0.107 0.226 0.762 -0.65 0.39 -0.63 0.17G351 0.017 0.864 0.085 0.246 0.720 -0.63 0.38 -0.56 0.22ALL 0.032 0.702 0.233 0.220 0.746 -0.54 0.43 -0.50 0.27


Fig. 1.The positions of the observed fields overplotted on an image of M31.


Fig. 2. Photometric errors as a function of V and I magnitude. The lines represent the treshold for three times the average error.The points above the lines are excluded from the final sample.


Fig. 3.The manually removed contaminating sources (open squares)from the entire sample superposed on the Color MagnitudeDiagram of the field G64. Most of the sources redder than (V − I ) ∼ 1.5 are galaxies or spurious stars from the decompositionof extended galaxies, most of the sources bluer than this figure are spurious stars from spikes and/or coronae of heavily saturatedstars


Fig. 4. Comparison between the photometry of the stars in common in the fields G287a and G287b. In the upper two panels thedifference in the final calibrated magnitudes is shown for the three WF cameras independently (first panel: V magnitudes, secondpanel: I magnitudes). In the lower two panels are reported the differences between the errors as computed by DoPHOT and theerror on the mean obtained by the two repeated measures of thestars in common, versus magnitude in the corresponding filter(third panel: V, fourth panel I. The differences has been averaged over 0.5 mag boxes. The errors provided by DoPHOT are goodestimates of the true photometric uncertainties.


Fig. 5. The (V,V-I) Color Magnitude Diagrams for the fields: G287, G87, G33, G322, G76, G119, G272, G108. In this figureand in figure 6 the diagrams are shown in order of increasing distance from the major axis, with the only exception of G327 (seetext).


Fig. 6. The (V,V-I) Color Magnitude Diagrams for the fields: G64, G58, G105, G11, G351, G219, G319, G327. See the captionof Fig. 5.


Fig. 7.The CMD of the G76 field represented as an isodensity contour plot (Hess diagram) for greater clarity. The main featuresdescribed in the text are indicated. The outermost countourcorrespond to a density of 10 stars per 0.1 mag× 0.1 magbox. Thestep between subsequent contours is 5 up to the 8th contour. The 9th contour corresponds to 75 stars per 0.1 mag× 0.1 magboxand after this contour the step is 100. The innermost contouris at 2200 stars per 0.1 mag× 0.1 mag, while the box sampling thepeak of the RHB Clump has a density of 2334 stars.


Fig. 8.CMD for the coadded halo fields zoomed to show the details of the HB morphology (see text). The lines superimposed tothe plot are the HB fiducial for M68, shifted by∆V=9.44 and∆V-I=0.09 to match the blue HB of the M31 fields, and the RGBridge lines of the M31 globular clusters G11 ([Fe/H] = -1.89) and G58 ([Fe/H] = -0.57). Panel (a) the CMDs of G105, G327,G319 and G219. Stars from different fields are marked with different symbols (G105: pentagons; G327: circles; G319: triangles;G219: squares). Panel (b) the CMDs of G11 and G351 (G11:×s; G351: stars). The BHB of the CMDs presented in panel (b) iscontaminated by an apparent blue plume that is not present inthe CMDs of panel (a).


Fig. 9.CMD of all the 154784 stars from the disk-dominated fields G76, G119 and G322, reported to the absolute plane [MI , (V−I )0]. Isochrones at [Fe/H] = 0.0 andY = 0.28, from the set of Bertelli et al. (1994) are superimposed onthe diagram. Thecontinuous lines are isochrones of age t= 60, 100, 200 and 400 Myr, from top to bottom. Open squares and circles correspondto isochrones of t= 1 and 12 Gyr, respectively. The stars brighter and fainter than theMI = −2 threshold are shown as points ofdifferent thickness to allow easier recognition of both the densely populated features in the lower part of the CMD (e.g. the HBClump) and the sparse bright features (e.g. the upper MS and the red plume of RSG stars


Fig. 10.The population boxes defined to study the YMS are evidenced inthe [V,(V-I)] CMD of the G64 field. The boxes markedwith Yu andYd samples different part of the Main Sequence, while the box marked withC has been defined for normalizationpurposes. The stars falling into the indicated boxes has been indicated with open circles.


Fig. 11. YMS (young main sequence) indices versus deprojected distance from the center of the galaxy (panels a and b), andversus absolute distance along the major axis (panel c). Where no error bars are seen, they are smaller than the symbol size.Only the fields with a substantial YMS population (e.g. disk fields) are plotted. The upper and lower panels plot the total fractionof stars younger than∼ 0.5 Gyr as a function of distance from the nucleus, while the center panel shows the fraction of youngstars as a function of galactocentric distance. Note that while the G76 field has the strongest young population, it is thesparselypopulated G87 field that has the youngest main sequence. The G76 field coincides with a ring of general enhanced star formation10 kpc from the nucleus.


Fig. 12.CMDs of the upper RGB in the absolute plane, for four fields, illustrating the full range of crowding in the data set. Theridge lines of template galactic globular clusters are superimposed to each plot. From left to right (CG metallicity scale): NGC6341 ([Fe/H] = −2.16), NGC 6205 ([Fe/H] = −1.39), NGC 5904 ([Fe/H] = −1.11) and NGC 104 ([Fe/H] = −0.70), NGC6553 ([Fe/H] = −0.16) and NGC 6528 ([Fe/H] = −0.07). The inner frame encloses the stars whose metallicitiesare determinedusing the interpolating scheme described in Fig. 13.


Fig. 13.This plot illustrates how our interpolation scheme works. To the left of jagged boundary, we interpolate in color, whileto the right, metallicity is determined using an interpolation in magnitude. Stars falling below the curved locus of NGC6528 areexcluded from the sample, but represent only≈ 1% of the total in any field.


Fig. 14. Histograms of the metallicity distributions, in the Carretta & Gratton (1997) scale, for the fields: G287, G87, G33, G322,G76, G119, G272 and G108. In the upper left corner of each panel are reported: the name of the field, the number of stars usedto derive the MD, the average metallicity ([Fe/H]ave) together with the associated standard deviation, the median metallicitytogether with the associated semi-interquartile interval, and the fractions of stars with: [Fe/H] < −1.4 [Metal-Poor or Young–MPor Y], −1.4 < [Fe/H] < −0.2 [Metal-Rich –MR], [Fe/H] > −0.2 [Very Metal-Rich – VMR], respectively.


Fig. 15. The same as Fig. 14 for the fields: G64, G58, G105, G11, G351, G219, G319 and G327.


Fig. 16.The metallicity distribution (same as Fig. 14, 15) for the field adjacent to the remote globular cluster G1, 35 kpc fromthe M31 nucleus. The frames from Rich et al. (1996) have been reduced in the same way as all other fields in our dataset.


Fig. 17.The metallicity distribution for the field G64 shows the effect of choosing different parameters. The detailed shape of theabundance distribution and the possible existence of a metal poor peak are both sensitive to these choices. Panel (a): Using ouradopted CG metallicity scale, distance modulus, and reddening. Panel (b): same as (a), ZW metallicity scale. Panel (c):(m−M)0

increased by 0.15 mag, and (d): (m− M)0 decreased by 0.15 mag. Panel (e):E(B− V) increased by 0.05 mag, and (f):E(B− V)decreased by 0.05 mag.


Fig. 18. Histograms of the metallicity distributions, in the Zinn & West (1984) scale, for the fields: G287, G87, G33, G322,G76, G119, G272 and G108. In the upper left corner of each panel are reported: the name of the field, the number of stars usedto derive the MD, the average metallicity ([Fe/H]ave) together with the associated standard deviation, the median metallicitytogether with the associated semi-interquartile interval, and the fractions of stars with: [Fe/H] < −1.6 [Metal-Poor or Young–MPor Y], −1.6 < [Fe/H] < −0.4 [Metal-Rich –MR], [Fe/H] > −0.4 [Very Metal-Rich – VMR], respectively.


Fig. 19. The same as Fig. 18 for the fields: G64, G58, G105, G11, G351, G219, G319 and G327.


Fig. 20. Map of the spatial and chemical substructures as described by Ferguson et al. (2002) (their Fig. 7). The positions ofour fields are plotted over the cartoon describing the location of the possible projected orbit of the ”giant stellar stream”, the”northern spur”, and the ”G1 clump”.


Fig. 21. Metallicity parameters, derived from the MDs, are plotted versus the angular distance from the center of M31. [Fe/H] <−1.4 [Metal-Poor or Young – MP or Y], and [Fe/H] > −0.2 [Very Metal-Rich – VMR], versus (a) the X-coordinate, (b) Y, (c) R,the galactocentric distance, respectively. A few fields areevidenced by different symbols (see Sect. 9.2.1): G287 (solid circle);G87 (solid square); G219 (star) and G319 (×).


Fig. 22.Panel (a): Fraction of the Metal-Poor Population (with [Fe/H] < −0.8 [MPP] plotted versus the absolute distance fromthe major axis|Y|. Panel (b): Median metallicity (GC97-scale) plotted vs. the absolute distance from the major axis|Y|.

arXiv:astro-ph/0212531v1 23 Dec 2002

Documents

Transcript of arXiv:astro-ph/0212531v1 23 Dec 2002